Optimization Algorithm of Tourism Security Early Warning ...

Research ArticleOptimization Algorithm of Tourism Security Early WarningInformation System Based on Long Short-Term Memory (LSTM)

Lei Feng1 and Yukai Hao 12

1School of Management Wuhan University of Technology Wuhan 400070 China2School of Economics and Management Tibet University Lhasa 850000 China

Correspondence should be addressed to Yukai Hao haoyukaiwhuteducn

Received 20 July 2021 Revised 18 August 2021 Accepted 20 August 2021 Published 8 September 2021

Academic Editor Syed Hassan Ahmed

Copyright copy 2021 Lei Feng and Yukai Hao )is is an open access article distributed under the Creative Commons AttributionLicense which permits unrestricted use distribution and reproduction in any medium provided the original work isproperly cited

Tourism safety is the focus of the tourism industry It is not only related to the safety of touristsrsquo lives and property but also relatedto social stability and sustainable development of the tourism industry However the security early warning of many scenic spotsfocuses on the response measures and remedial plans after the occurrence of security incidents and the staff of many scenic spotshave limited security awareness and information analysis ability which is prone to lag in information release and do not payattention to the information of potential security problems )erefore this paper studies the optimization algorithm of thetourism security early warning information system based on the LSTMmodel and uses the recurrent neural network and LSTM toimprove the processing and prediction ability of time-series data )e experimental results show that the number of three hiddenlayers in the tourism security early warning information system based on the LSTM model can reduce the training time of themodel and improve the performance Compared with the tourism safety early warning information system based on the BP neuralnetwork it has better accuracy and stability has better processing and prediction ability for time series data and can monitor andanalyze data scientifically in real-time and dynamically analyze data

1 Introduction

Tourism has gradually become one of the important in-dustries As people no longer meet the basic needs of lifemore and more people begin to pursue high-quality lifeTourism has gradually become one of the important in-dustries More and more people begin to pursue tourism life[1] In addition the tourism resources are constantly de-veloped and utilized and the tourism environment andcontent are constantly changing In recent years in additionto characteristic cultural city tourism there are also naturaland cultural tourism marathon competition in naturaltourism areas etc )ese emerging tourism projects not onlyattract more tourists but also improve the economic growthof the tourism industry [2 3] However there are manyunfortunate events that tourists encounter in the process oftourism such as abrupt weather changes in marathoncompetitions in natural scenic spots stampede on the Bund

of Shanghai caused by too many tourists tsunami kid-napping of tourists and constant theft in tourist areas whichhave caused serious adverse effects on the tourism industryand restricted the sustainable development of the tourismindustry [4 5] )erefore tourism security has become ahighly valued and concerned issue in various countries andregions and tourism security early warning has become aninevitable trend of tourism development With the devel-opment and application of intelligent information tech-nology many scenic spots will collect and release the safetyinformation and early warning information of scenic spotsthrough intelligent wearable products and correspondingapps based on big data such as the information of dangerousareas of scenic spots and the number of visitors to scenicspots Although the danger that some tourists will encounterin the scenic spots can be avoided to a certain extent thereare some problems in many scenic spots such as untimelyrelease of safety early warning information low safety

HindawiComputational Intelligence and NeuroscienceVolume 2021 Article ID 9984003 11 pageshttpsdoiorg10115520219984003

awareness of staff and error in judgment of correspondinginformation At the same time the focus of many tourismsafety measures implementation plans and methods inscenic spots is that after the occurrence of tourism safetyevents the corresponding information release channel isnarrow and there is a lack of relevant knowledge reserve andmature and stable response plan in advance warning [6])is shows that the development of tourism security earlywarning in the tourism industry can no longer meet theneeds of the development of the tourism industry)ereforethe tourism security early warning information system thatthe tourism industry needs to build can conduct accurateand scientific information analysis on the collected securityinformation of relevant scenic spots in real-time and ef-fectively and output the information analysis results in timeand improve the efficiency of safety early warning infor-mation in scenic spots

)is paper studies the optimization algorithm of thetourism security early warning information system based onthe LSTM model Compared with the traditional tourismsecurity early warning methods the artificial neural networkhas better fault tolerance and stronger robustness It canquickly process data and find the corresponding optimalsolution and its nonlinear thinking can well deal with therelationship between many factors Compared with the BPneural network the LSTM model can better process tem-poral information and realize the purpose of real-timeprocessing tourism safety early warning information )ispaper is mainly divided into the following three parts )efirst part introduces the development and related concepts oftourism security early warning information system and thedevelopment and application of the LSMT recurrent neuralnetwork)e second part constructs a tourism early warninginformation system based on the BP neural network andintroduces the recursive neural network and LSTM to op-timize the algorithm of the tourism early warning infor-mation system In addition the corresponding tourismsecurity early warning information indicators are con-structed by integrating various factors of tourism security Inthe third part the optimization algorithm of the tourismsecurity early warning information system based on theLSTMmodel is trained and tested and the simulation resultsare analyzed

2 Related Work

)e tourism security early warning information systemcontains a complex system of many influencing factorswhich can evaluate and analyze various security indicators oftourism destination and determine the change trend of thesystem composed of the overall environment of tourismdestination so as to early warn and eliminate the securityincidents that may occur in the security system [7 8] )etourism safety early warning information system can pro-mote the sustainable development of tourism destinationand improve the satisfaction of touristsrsquo experience andpersonal safety and has important guiding significance in thelong-term development of tourism industry and socialeconomy natural environment and social stability [9]

)erefore the research of the tourism security early warninginformation system has always been the focus of attentionTourism safety factors are diversified and their externalmanifestations can be roughly divided into natural disastersdiseases crimes traffic safety and others Many of them areuncontrollable but scenic spots can still analyze some po-tential risk factors according to the analysis of relevantinformation Some scholars have proposed an IntelligentTourism early warning system for the stampede in scenicspots that is to analyze and guide the monitored datathrough intelligent services and processing functions [10])is method is more suitable for use in urban scenic spotsand its early warning focuses on the tourism safety problemscaused by human factors According to the characteristics ofnatural scenic spots some scholars proposed to establish therisk identification and evaluation model of natural scenicspots through the combination of the GIS and Bayesiannetwork model [11] )is method has strong pertinence andcan clarify the scope of risk and improve the accuracy oftourism safety early warning but it needs long-term effectivedata as the basis of decision-making which greatly increasesthe time cost Some scholars proposed to build a safety earlywarning system based on the BP neural network Itsmodeling is relatively simple and can obtain informationanalysis results in a relatively short time [12] However theBP neural network is weak in the analysis of time seriesinformation and its output early warning results tend tostatic analysis And with the increase of the types of riskfactors its accuracy is also affected to a certain extent Inaddition some research on tourism security early warningmostly focuses on the application mechanism of the artificialneural network in tourism security emergencies whichprovides theoretical support and lays a solid theoreticalfoundation for tourism security research [13] In additionaccording to the current situation of tourism environmentresearchers put forward to explore the ecological deterio-ration and sudden environmental security problems causedby tourism activities from an ecological perspective predictthe ecological environment status of tourism destinationsand make targeted preventive measures and rescue plans[14] However from the aspect of tourism security crisisearly warning and management the tourism security earlywarning system based on the BP neural network still hasmany deficiencies in processing time series data and needs tobe further optimized

)e main objects of tourism security early warningsystem are tourists or local residents [15 16] )erefore theinformation it provides is more detailed which has a goodeffect in the security of outbound tourism However sometravellers ignore early warning information or do not payattention to relevant early warning information in time andthey do not pay attention to early warning information andsuggestions [17] It should be noted that the tourism safetyearly warning information system does not specifically es-tablish a long-term safety early warning information systemfor tourism but is issued by the relevant meteorologicalbureau and the Safety Supervision Bureau and other de-partments carry out classified early warning for naturaldisasters and social security events and the information

2 Computational Intelligence and Neuroscience

subject is not limited to tourists [18 19] A similar tourismsecurity early warning information system has been estab-lished and the warning language is relatively mild )ispaper briefly introduces the tourism destination countriesbut it does not clearly classify the contents of early warning[20] At the same time citizenship does not connect tourismservices so few people pay attention to the released tourismsecurity early warning information [21 22] Researchershave been constantly trying and studying hoping to build amore scientific tourism security early warning informationsystem [23]

21 Construction and Optimization of Tourism Security EarlyWarning Information System Based on LSTM )e tourismsafety early warning information system is used to predictand warn the changes of scenic spots in the future frommultiple dimensions according to the reasonable indexsystem and scientific methods )erefore the influencingfactors of tourism safety early warning information arediversified and nonlinear In this paper the LSTM model isused to construct the tourism safety early warning infor-mation system which improves the processing ability of thesystem to temporal information so as to realize the purposeof real-time dynamic information supervision and analysisFigure 1 shows the flowchart of the tourism security earlywarning information system based on the LSTM model

22 Construction of Tourism Safety Early Warning NeuralNetwork Information System )e artificial neural networkwhich simulates the connection between human neuronscan process the relevant signals obtain the data signalprediction model and solve the nonlinear data predictionand other related problems )erefore this paper selects theBP neural network as the foundation of the tourism securityearly warning information system and extracts the implicitrelationship of the static data that need to be analyzed andpredicted [24] )e neurons of the BP neural network canconnect multiple inputs but only have one output node asshown in Figure 2

)e input layer of the multilayer perceptron is repre-sented as Lin and x (x1 x2 x3)

T the hidden layer as Lhiddenand h (h1 h2 h3 h4)

T and the output layer as Lout andy (y1 y2 y3) )e process of data transmission frominput layer to output layer is forward propagation as shownin

h f W1T

x + b1

1113872 1113873 (1)

y f W2T

x + b2

1113872 1113873 (2)

)e weight matrix of data transfer is expressed asW1 isin R3times4 the weight of data transfer connection betweenhidden layer and output layer is expressed as W2 isin R4times3and the bias of output layer is expressed as b2 isin R3times1 )eactivation function f is shown in

f 1

1 + exp(minusx) (3)

In the learning process of the BP neural network theweights and thresholds are modified by the gradient method)e iterative formulas after the rest are shown in

ΔW(n + 1) minusηzE

zW(n)+ a middot ΔW(n)

Δθ(n + 1) minusηzE

zθ(n)+ a middot Δθ(n)

⎧⎪⎪⎪⎪⎪⎪⎨

⎪⎪⎪⎪⎪⎪⎩

(4)

ΔW(n + 1) W(n + 1) minus W(n)

Δθ(n + 1) θ(n + 1) minus θ(n)1113896 (5)

Although the BP neural network can solve the non-linear problem the data processed by the BP neuralnetwork have no correlation on the time line that is itcannot process the time series characteristic data relatedto the last moment or earlier data An LSTM is designedbased on the structure of the BP neural network )estructure of function of network memory that is it caneffectively remember the data characteristics of time di-mension in training data and its structure is shown inFigure 3

)e difference between the LSTM and BP neuralnetwork is that the nodes between the hidden layers areinterconnected so that the input of the hidden layer atthat time contains two parts that is the output trans-mitted by the input layer at this time and the output of thehidden layer at the last moment so that the hidden layercontains the information memory at this time and at thelast moment or earlier [25] )e transfer process is shownin

ht f W1T

xt + WhT

htminus1 + b1

1113872 1113873 (6)

yt f W2T

ht + b2

1113872 1113873 (7)

And b2 represents the output layer offset When the timeis t the input is xt the hidden layer output is ht and theoutput is yt when the time is t minus 1 the hidden layer output ishtminus1 )e activation function formula is shown in

tanhx sinhx

cosh x

ex

minus eminusx

ex

+ eminusx (8)

)e sum of the losses is the total loss function Let theloss function be labeled as the negative log likelihoodfunction as shown in

Lt

minusytlog o

t (9)

)en the total loss function of the sequence is shown in

L x1 x

t1113966 1113967 y

1 y

t1113966 11139671113872 1113873 1113944 L

t (10)

)e development of LSTM can be transformed into thecorresponding BP neural network and trained by theBPTT algorithm [26 27] If it is necessary to train t timedata the LSTM is expanded into a BP neural network witht hidden layer It can be seen from Figure 2 that theparameters in the same position after the expansion of the

Computational Intelligence and Neuroscience 3

LSTM are shared with each other while the BP neuralnetwork is not shared so the LSTM greatly reduces thelearning and training parameters According to the rel-evant theory it can be considered that the length of thesequence data that can be processed by LSTM is infiniteand the hidden layer of processing information is alsoinfinite )erefore in the actual information transmissionprocess the information in the hidden layer neurons willgradually weaken and lose due to the extension of timethat is the gradient vanishing phenomenon )is leads tothe decrease of prediction performance In addition thelength of the delay window needs to be determined inadvance when the LSTM is trained which improves thedifficulty of automatically obtaining the optimal param-eters in practical application

23 Optimization of Tourism Security Early Warning Infor-mation System Based on LSTM In order to solve the aboveproblems of LSTM so as to make its memory long-termFigure 4 shows the LSTM structure development diagram

It can be seen from the figure that the LSTM neuralnetwork introduces controllable so as to avoid the gradientdisappearance problem of LSTM )ere are four layers inLSTM cell including forgetting gate input layer outputlayer and update layer which interact with each other in aspecial way And we provide corresponding continuouswriting reading and reset functions )at is to say theLSTM neural network adds a C state for long-term infor-mation memory on the basis of the recurrent neural net-work which is the unit structure of LSTM If the time at thistime is t when the forgetting gate layer controls the amount

Tourism security early warning informationindex system

Standardizedtreatment

LSTM neural network

Evaluation of LSTM neural networkmodel

Warning range

Alarm area Vigilancezone

Pre control countermeasures toeliminate police and patients

Safety zone

Early warning indexcalculation

Prediction of earlywarning index

LSTM recurrent neural network evaluationof tourism security information

move incircles

move incircles

Trained LSTM neuralnetwork


observed datahistorical data

Figure 1 Flowchart of the tourism security early warning information system based on the LSTM model


Lin

Lhidden

Lout

x

h

y

x tndash2 x tndash1 x

h tndash2

y tndash2

h tndash1

y tndash1

h

y

Delayspread

W1b1

W2b2

WhW1b1

W2b2

W1b1

W2b2

W1b1

W2b2

WhWhWh

Figure 3 )e structure of network memory function

X +

X X

tabhδ δ δ

tanh

xt

ht

A A

xtndash1 xt+1

htndash1 ht+1

Figure 4 LSTM structure expansion

Lin Lhidden Lout

x1

x3

h3

y1

x2

h2

h4

y3

h1

y2

W1b1 W2b2

Figure 2 BP neural network simple model


of information transferred from the previous unit state ctminus1to the current ct state

ft σ wf lowast htminus1 xtminus11113858 1113859 + bf1113872 1113873 (11)

)e main purpose of the input gate is to filter infor-mation to avoid useless information entering the current ct

state )e sigmoid layer and tanh layer of the input gate canupdate the state )e formulas are shown in

it σ wi + htminus1 xt1113858 1113859 + bi( 1113857 (12)

gt tanh wc lowast htminus1 xt1113858 1113859 + bc( 1113857 (13)

where σ is sigmoid function and the numerical range is[0 1] After that the last moment state ctminus1 and ft aremultiplied to update the state )e useless information isfiltered out and a new value it lowastgt is added )e corre-sponding adjustment is made according to the individualupdate state )e formula is shown in

ct ft lowast ctminus1 + it lowastgt (14)

)e output gate is used to control the current outputaffected by long-term storage information It mainly de-termines the cell to be output through sigmoid layer thensets the cell in [minus1 1] range by tanh and multiplies thecorresponding output gate Finally the determined outputpart is output as shown in

ot σ w0 lowast htminus1 xt1113858 1113859 + b0( 1113857 (15)

ht ot lowast tanh ct( 1113857 (16)

It can be seen from the above formula that sigmoidfunction is the activation function of input output andforgetting gates with values in [0 1] )e other activationfunction tanh as shown in formula (6) is also commonlyused in the input and output gates of LSTM and itsmonotonicity is more consistent with the characteristics ofneurons in neural networks Figure 5 shows two kinds ofactivation function diagrams

24 Index Construction of Tourism Security Early WarningInformation System Tourism safety includes many influ-encing factors According to the relevant analysis and in-duction the index of tourism early warning informationsystem in this paper is three levels that is the first level istourism safety early warning and the second level has fourindicators that is the stability of tourism natural disastersthe safety of Tourism travel facilities the safety of tourismdestinations and the safety of tourism environmental pol-lution In addition each first level indicator also containsthree levels of impact factors

Tourism safety early warning mode is divided intoexcellent level good level qualified level and critical levelAmong them the excellent level indicates that the overallenvironment of the tourism destination has high securitythere is no hidden danger and there is no need to worryabout the occurrence of emergencies Good level means

that the overall environment of the tourism destinationhas high security Although there may be potential safetyhazards and the possibility of small-scale security emer-gencies the probability of occurrence is very small andthere are sound and mature treatment plans and remedialmeasures From the perspective of realistic probabilitypotential tourism safety accidents may occur Howeverthe impact of such accidents can be effectively controlledwithin the corresponding range and there is a good re-sponse plan which requires tourists to have a certaindegree of cognition and knowledge of potential safetyhazards Tourists who do not have this condition are notencouraged to enter the tourist destination )e criticallevel means that the tourism destination has a highprobability of serious tourism safety accidents and be-cause there is no corresponding treatment plan andmeasures once a safety accident occurs it will have se-rious or even catastrophic consequences for tourists andthe tourism destination Tourists should be preventedfrom entering the tourism destination within this level Inorder to show the four tourism safety early warning modesmore intuitively the four level alarms are matched withcorresponding early warning signals that is the safetylevel early warning signal is green the good level earlywarning signal is blue the qualified level early warningsignal is orange and the critical level signal light is redFigure 6 shows the warning value and discriminationmode of tourism security early warning

100

075

050

025

000-4 -2 0

Sigmoid

2 4

1

05

0

ndash05

ndash1-4 -2 0

Tanh

2 4

Figure 5 Two activation function diagrams


25 Test Results of Tourism Security Early WarningInformation System Based on LSTM

251 Optimization Test Results of Tourism Security EarlyWarning Information System Based on LSTM In the LSTMalgorithm the time step represents the length of the indexsequence that can be used which has a certain impact on themodel )erefore under the condition that the algorithmremains unchanged the performance of the algorithm withthe step size of 4 45 and 90 is tested as shown in Figure 7

It can be seen from the results in the figure that theLSTM algorithm will continuously improve the corre-sponding prediction performance with the increase of timestep When the time step increases to a certain length theaccuracy of LSTM algorithm decreases In addition the timestep can reflect the correlation length of the data in the timeseries If the time step is too short the correlation infor-mation between the data will be insufficient which willreduce the prediction effect of the algorithm When the steplength is too long it will reduce the correlation between thedata because of too much redundant data thus reducing theprediction accuracy of the algorithm so the selection of thestep length algorithm is very important

According to the LSTM recurrent algorithm the in-crease of the number of layers will improve the learning

performance but layers will also lead to the improvement ofthe complexity of the algorithm system affect its conver-gence speed consume more time in the sample training andincrease the difficulty of training )erefore this paper tests

Warning value and discriminant model of tourism safety

Tourismfacilities

usesaturation

080- 095 070- 1 00 0 00- 010 0 00- 010 000- 005 070- 100 0 45- 1 00 0 00- 010 000- 005 0 00- 010 Securitylevel

070- 080 060- 0 70 0 10- 015 0 10- 015 005- 006 060- 0 70 0 35- 0 45 0 10- 020 005- 010 0 10- 020 A goodlevel

060- 070 050- 0 60 0 15- 020 0 15- 020 006- 008 050- 0 60 0 25- 0 35 0 20- 030 010- 020 0 20- 030 qualified

000- 060 000- 050 0 20- 100 0 20- 100 gt008 000- 050 0 00- 0 25 0 30- 100 020- 100 0 30- 100 Criticallevel

Politicalstability

Frequencyof

occurrenceof

hydrometeorologicaldisasters

Frequencyof

earthquakeand

geologicaldisasters

Realunemployment rate

Socialsecuritystability

Trafficsafety

Theconsumer

price indexrose

Thefrequency

ofoutbreaks

ofepidemicdiseases

Potentialindex ofculturalconflictbetweenhost and

guest

Alarmoutput

indicatingsignal

Figure 6 Warning value and discriminant model of tourism safety

06

05

04

03

02

01

00 200 400

The number of iterations

The time step is equal to 4The time step is equal to 45The time step is equal to 90

Root

mea

n sq

uare

erro

r

600 800

Figure 7 LSTM recursive neural network algorithm does notsynchronize the long performance test


the performance of the LSTM algorithm as shown inFigure 8

In the figure that the convergence effect of LSTMimproves with the increase of layers but the corre-sponding training and testing time is also longer andlonger And when the number of layers of LSTM increasesto four the improvement of its performance is not ob-vious but it takes a long time )erefore considering theneeds of all aspects the three-layer LSTM algorithm is themost appropriate As shown in Figure 9 the hidden layerof LSTM algorithm contains different numbers of nodeloss function values

It can be seen from the figure that when the number ofhidden layer nodes reaches 520 the loss function value of theLSTM algorithm reaches the minimum Compared with theloss function of the hidden layer with 130 and 260 nodes itcan be seen that with the increase of the number of nodesthe corresponding loss function value decreases signifi-cantly )is shows that when the number of hidden layer

nodes is large enough the fitting performance of the LSTMalgorithm can be brought into full play

26 Simulation Test Results of Tourism Security EarlyWarning Information System Based on LSTM As shown inFigure 10 it is the error comparison chart of the BP neuralnetwork algorithm and LSTM algorithm for tourism safetyindex prediction results In the results of the figure theprediction result of the LSTM algorithm is closer to the realvalue)e algorithm has a large error in the prediction resultsof individual values mainly because the BP neural network isprone to the problem of local optimal solution )erefore interms of accuracy and stability the prediction accuracy ofLSTM for time series data is higher and the stability is better

As shown in Figure 11 it is an early warning analysismodel information system based on LSTM Tourism des-tination security is a complex dynamic change so the inputvalue of the tourism security early warning informationsystem can be not only discrete variables but also continuous

218161412

Root

mea

n sq

uare

erro

r

108060402

00 200 400

The number of iterations600 800

2 layer LSTM neural network3 layer LSTM neural network4 layer LSTM neural network

Figure 8 LSTM recursive neural network algorithm was used to test the level dependent performance

09

08

07

06

05

04

Loss

func

tion

valu

e

03

02

01

00 5 10 15

Number of training

Number of hidden layer nodes130Number of hidden layer nodes260Number of hidden layer nodes520Number of hidden layer nodes1040

20 25 30

Figure 9 )e hidden layer of the LSTM recursive neural networkalgorithm contains different numbers of node loss function values

15

1

05

0

Erro

r dev

iatio

n

ndash05

ndash1230 250 270 290 310 330

e serial number

BP neural networkLSTM neural network

350 370 390 410 430 450 470 490 510

Figure 10 )e error comparison graph of the BP neural networkalgorithm and LSTM recursive neural network algorithm fortourism safety index prediction results


variables and the output value belongs to Boolean discretevector

)e security status of tourism destination is divided intodifferent levels and the output value information systembased on LSTM is set as a vector between 0 and 1 When them-th index element represents 1 and the other index ele-ments represent 0 the security of tourism destination is in acertain level )e simulation test results the tourism securityearly warning information system as shown in Table 1

3 Conclusion

With the continuous development of economy people beginto see the difference of the world through tourism on thebasis of meeting the basic life However what is not matchedwith the booming tourism industry is the tourism securityearly warning information system Tourism security is acomprehensive problem composed of many factors which isnot only related to the life and property safety of tourists butalso related to social stability and the development andprotection of tourism resources At present the tourism isrelatively backward focusing on the remedial measures andtreatment after the occurrence of security incidents whichcannot play the role of early warning to reduce disaster

losses )erefore this paper studies the optimization algo-rithm of the tourism security early warning informationsystem based on LSTM On the basis of the tourism securitybased on the BP neural network it uses recurrent neuralnetwork and LSTM to optimize the system algorithm so asto improve the ability of the early warning informationsystem to process and predict the time series data )eexperimental results show that the learning ability andconvergence effect of LSTM model will improve with theincrease of the number of hidden layers but when it in-creases to a certain number the increase of learning abilityand convergence effect is not obvious )erefore it isnecessary to set an appropriate number of hidden layers forthe LSTM model to improve its performance )e tourismsecurity early warning information system based on theLSTM model has better accuracy and stability than thetourism security early warning information system based onthe BP neural network algorithm has better processing andprediction ability for time series data and is more in linewith the needs of the tourism security early warning in-formation system In addition compared with othermethods the tourism security early warning informationsystem based on the LSTM model can be applied to a widerrange whether it is a tourist city scenic spot or a tourist

Warninginstructions

Thresholdvalue range

Tourism resources arenot fully utilized

Security transformationof tourist destinations

Police line on

The p line

Time

Figure 11 An early warning analysis model of the tourism security early warning information system based on the LSTM recursive neuralnetwork

Table 1 Based on LSTM recursive neural network tourism security early warning information system simulation test results

X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 Alarm indication1 095 074 0089 006 002 072 044 004 003 012 (1000) green2 087 061 006 011 005 067 036 015 006 011 (1000) green3 074 066 014 016 008 067 044 020 009 016 (0100) blue4 067 051 016 017 007 054 034 028 012 032 (0010) orange5 082 067 008 011 005 068 039 015 006 014 (0100) blue6 058 058 016 020 009 050 027 031 015 022 (0010) orange7 062 048 027 020 012 050 025 036 022 032 (0001) red8 057 041 021 022 015 042 028 044 027 051 (0001) red


natural scenic spot or it can be combined with intelligentwearable devices for data collection and analysis Howeverthe experimental data in this paper are mainly for theanalysis of the indicators of the scenic spot so the indexsystem needs to be further improved In the future devel-opment the tourism safety early warning informationsystem between scenic spots should be connected with eachother to strengthen the information circulation At the sametime set up a tourism safety early warning informationsubsystem for economically underdeveloped scenic spots toreduce the cost of tourism safety early warning informationsystem on the basis of ensuring the safety of scenic spots andtourists

Data Availability

)e data used to support the findings of this study areavailable from the corresponding author upon request

Conflicts of Interest

)e authors declare that they have no conflicts of interestregarding the publication of this paper

Acknowledgments

)is work was supported by the Ministry of Science andTechnology of the Peoplersquos Republic of China projectnumber 2018YFB0804300

References

[1] H Benbrahim H Hachimi and A Amine ldquoDeep transferlearning with Apache spark to detect covid-19 in chest x-rayimagesrdquo Romanian Journal of Information Science andTechnology vol 23 no S pp S117ndashS129 2020

[2] P Zhao Y Liu H Liu and S Yao ldquoA sketch recognitionmethod based on deep convolutional-recurrent neural net-workrdquo Journal of Computer-Aided Design amp ComputerGraphics vol 30 no 2 pp 217ndash224 2018

[3] R McKinley R Wepfer F Aschwanden et al ldquoSimultaneouslesion and brain segmentation in multiple sclerosis using deepneural networksrdquo Scientific Reports vol 11 no 1pp 1087ndash1111 2021

[4] R A Bhuiyan S Tarek and H Tian ldquoEnhanced bag-of-wordsrepresentation for human activity recognition using mobilesensor datardquo Signal Image and Video Processing vol 2021Article ID 1907-4 8 pages 2021

[5] H N Dai H Wang G Xu J Wan and M Imran ldquoBig dataanalytics for manufacturing internet of things opportunitieschallenges and enabling technologiesrdquo Enterprise InformationSystems vol 14 no 9-10 pp 1279ndash1303 2020

[6] G Gui Z Zhou J Wang F Liu and J Sun ldquoMachinelearning aided air traffic flow analysis based on aviation bigdatardquo IEEE Transactions on Vehicular Technology vol 69no 5 pp 4817ndash4826 2020

[7] B M H Abidine L Fergani B Fergani and M Oussalahldquo)e joint use of sequence features combination and modifiedweighted SVM for improving daily activity recognitionrdquoPattern Analysis amp Applications vol 21 no 1 pp 119ndash1382018

[8] S Wan L Qi X Xu C Tong and Z Gu ldquoDeep learningmodels for real-time human activity recognition withsmartphonesrdquo Mobile Networks and Applications vol 25no 2 pp 743ndash755 2020

[9] A Amaya P P Biemer and D Kinyon ldquoTotal error in a big dataworld adapting the TSE framework to big datardquo Journal ofSurvey Statistics andMethodology vol 8 no 1 pp 89ndash119 2020

[10] F Y Zhou L P Jin and J Dong ldquoA review of convolutionalneural networksrdquo Chinese Journal of Computers vol 40 no 6pp 1229ndash1251 2017

[11] Z Xu C Cheng and V Sugumaran ldquoBig data analytics ofcrime prevention and control based on image processingupon cloud computingrdquo Journal of Surveillance Security andSafety vol 1 no 1 pp 16ndash33 2020

[12] B Shi X Bai and C Yao ldquoAn end-to-end trainable neuralnetwork for image-based sequence recognition and its ap-plication to scene text recognitionrdquo IEEE Transactions onPattern Analysis and Machine Intelligence vol 39 no 11pp 2298ndash2304 2017

[13] X Li Z Zhao and F Liu ldquoBig data assimilation to improvethe predictability of COVID-19rdquo Geography and Sustain-ability vol 1 no 4 pp 317ndash320 2020

[14] M Geist P Petersen M Raslan R Schneider andG Kutyniok ldquoNumerical solution of the parametric diffusionequation by deep neural networksrdquo Journal of ScientificComputing vol 88 no 1 pp 1ndash37 2021

[15] M Mathew D Karatzas and C V Jawahar ldquoDocvqa Adataset for vqa on document imagesrdquo in Proceedings of theIEEECVF Winter Conference on Applications of ComputerVision pp 2200ndash2209 Waikola HI USA August 2021

[16] A E R ElSaid J Karns Z Lyu D Krutz A Ororbia andT Desell ldquoImproving neuroevolutionary transfer learning ofdeep recurrent neural networks through network-aware ad-aptationrdquo in Proceedings of the 2020 Genetic and EvolutionaryComputation Conference pp 315ndash323 Prague Czech Re-public March 2020

[17] A Vernotte M Valja M Korman G Bjorkman M Ekstedtand R Lagerstrom ldquoLoad balancing of renewable energy acyber security analysisrdquo Energy Informatics vol 1 no 15 pages 2018

[18] M Kend and L A Nguyen ldquoBig data analytics and otheremerging technologies the impact on the Australian auditand assurance professionrdquo Australian Accounting Reviewvol 30 no 4 pp 269ndash282 2020

[19] J M Johnson and T M Khoshgoftaar ldquo)e effects of datasampling with deep learning and highly imbalanced big datardquoInformation Systems Frontiers vol 22 no 5 pp 1113ndash11312020

[20] A T Fadi and B D Deebak ldquoSeamless authentication forIoT-big data technologies in smart industrial applicationsystemsrdquo IEEE Transactions on Industrial Informatics vol 17no 4 pp 2919ndash2927 2020

[21] R Feng D Grana and N Balling ldquoUncertainty quantificationin fault detection using convolutional neural networksrdquoGeophysics vol 86 no 3 pp M41ndashM48 2021

[22] C Baskin N Liss E Schwartz et al ldquoUniq uniform noiseinjection for non-uniform quantization of neural networksrdquoACM Transactions on Computer Systems vol 37 no 1ndash4pp 1ndash15 2021

[23] A Skolik J R McClean M Mohseni P Van Der Smagt andM Leib ldquoLayerwise learning for quantum neural networksrdquoQuantum Machine Intelligence vol 3 no 1 pp 1ndash11 2021

[24] M Aboelmaged and S Mouakket ldquoInfluencing models anddeterminants in big data analytics research a bibliometric


analysisrdquo Information Processing amp Management vol 57no 4 Article ID 102234 2020

[25] J L Leevy and T M Khoshgoftaar ldquoA survey and analysis ofintrusion detectionmodels based on cse-cic-ids2018 big datardquoJournal of Big Data vol 7 no 1 pp 1ndash19 2020

[26] A Glowacz ldquoFault diagnosis of electric impact drills usingthermal imagingrdquo Measurement vol 171 Article ID 1088152021

[27] W Wang N Kumar J Chen et al ldquoRealizing the potential ofthe internet of things for smart tourism with 5G and AIrdquo IEEENetwork vol 34 no 6 pp 295ndash301 2020


awareness of staff and error in judgment of correspondinginformation At the same time the focus of many tourismsafety measures implementation plans and methods inscenic spots is that after the occurrence of tourism safetyevents the corresponding information release channel isnarrow and there is a lack of relevant knowledge reserve andmature and stable response plan in advance warning [6])is shows that the development of tourism security earlywarning in the tourism industry can no longer meet theneeds of the development of the tourism industry)ereforethe tourism security early warning information system thatthe tourism industry needs to build can conduct accurateand scientific information analysis on the collected securityinformation of relevant scenic spots in real-time and ef-fectively and output the information analysis results in timeand improve the efficiency of safety early warning infor-mation in scenic spots

)is paper studies the optimization algorithm of thetourism security early warning information system based onthe LSTM model Compared with the traditional tourismsecurity early warning methods the artificial neural networkhas better fault tolerance and stronger robustness It canquickly process data and find the corresponding optimalsolution and its nonlinear thinking can well deal with therelationship between many factors Compared with the BPneural network the LSTM model can better process tem-poral information and realize the purpose of real-timeprocessing tourism safety early warning information )ispaper is mainly divided into the following three parts )efirst part introduces the development and related concepts oftourism security early warning information system and thedevelopment and application of the LSMT recurrent neuralnetwork)e second part constructs a tourism early warninginformation system based on the BP neural network andintroduces the recursive neural network and LSTM to op-timize the algorithm of the tourism early warning infor-mation system In addition the corresponding tourismsecurity early warning information indicators are con-structed by integrating various factors of tourism security Inthe third part the optimization algorithm of the tourismsecurity early warning information system based on theLSTMmodel is trained and tested and the simulation resultsare analyzed

2 Related Work

)e tourism security early warning information systemcontains a complex system of many influencing factorswhich can evaluate and analyze various security indicators oftourism destination and determine the change trend of thesystem composed of the overall environment of tourismdestination so as to early warn and eliminate the securityincidents that may occur in the security system [7 8] )etourism safety early warning information system can pro-mote the sustainable development of tourism destinationand improve the satisfaction of touristsrsquo experience andpersonal safety and has important guiding significance in thelong-term development of tourism industry and socialeconomy natural environment and social stability [9]

)erefore the research of the tourism security early warninginformation system has always been the focus of attentionTourism safety factors are diversified and their externalmanifestations can be roughly divided into natural disastersdiseases crimes traffic safety and others Many of them areuncontrollable but scenic spots can still analyze some po-tential risk factors according to the analysis of relevantinformation Some scholars have proposed an IntelligentTourism early warning system for the stampede in scenicspots that is to analyze and guide the monitored datathrough intelligent services and processing functions [10])is method is more suitable for use in urban scenic spotsand its early warning focuses on the tourism safety problemscaused by human factors According to the characteristics ofnatural scenic spots some scholars proposed to establish therisk identification and evaluation model of natural scenicspots through the combination of the GIS and Bayesiannetwork model [11] )is method has strong pertinence andcan clarify the scope of risk and improve the accuracy oftourism safety early warning but it needs long-term effectivedata as the basis of decision-making which greatly increasesthe time cost Some scholars proposed to build a safety earlywarning system based on the BP neural network Itsmodeling is relatively simple and can obtain informationanalysis results in a relatively short time [12] However theBP neural network is weak in the analysis of time seriesinformation and its output early warning results tend tostatic analysis And with the increase of the types of riskfactors its accuracy is also affected to a certain extent Inaddition some research on tourism security early warningmostly focuses on the application mechanism of the artificialneural network in tourism security emergencies whichprovides theoretical support and lays a solid theoreticalfoundation for tourism security research [13] In additionaccording to the current situation of tourism environmentresearchers put forward to explore the ecological deterio-ration and sudden environmental security problems causedby tourism activities from an ecological perspective predictthe ecological environment status of tourism destinationsand make targeted preventive measures and rescue plans[14] However from the aspect of tourism security crisisearly warning and management the tourism security earlywarning system based on the BP neural network still hasmany deficiencies in processing time series data and needs tobe further optimized

)e main objects of tourism security early warningsystem are tourists or local residents [15 16] )erefore theinformation it provides is more detailed which has a goodeffect in the security of outbound tourism However sometravellers ignore early warning information or do not payattention to relevant early warning information in time andthey do not pay attention to early warning information andsuggestions [17] It should be noted that the tourism safetyearly warning information system does not specifically es-tablish a long-term safety early warning information systemfor tourism but is issued by the relevant meteorologicalbureau and the Safety Supervision Bureau and other de-partments carry out classified early warning for naturaldisasters and social security events and the information








h f W1T

x + b1

1113872 1113873 (1)

y f W2T

x + b2

1113872 1113873 (2)


f 1

1 + exp(minusx) (3)






⎧⎪⎪⎪⎪⎪⎪⎨

⎪⎪⎪⎪⎪⎪⎩

(4)


Δθ(n + 1) θ(n + 1) minus θ(n)1113896 (5)



ht f W1T

xt + WhT

htminus1 + b1

1113872 1113873 (6)

yt f W2T

ht + b2

1113872 1113873 (7)


tanhx sinhx

cosh x

ex

minus eminusx

ex

+ eminusx (8)


Lt

minusytlog o

t (9)


L x1 x

t1113966 1113967 y

1 y

t1113966 11139671113872 1113873 1113944 L

t (10)








LSTM neural network


Warning range



Safety zone




move incircles

move incircles






Lin

Lhidden

Lout

x

h

y


h tndash2

y tndash2

h tndash1

y tndash1

h

y

Delayspread

W1b1

W2b2

WhW1b1

W2b2

W1b1

W2b2

W1b1

W2b2

WhWhWh


X +

X X

tabhδ δ δ

tanh

xt

ht

A A

xtndash1 xt+1

htndash1 ht+1


Lin Lhidden Lout

x1

x3

h3

y1

x2

h2

h4

y3

h1

y2

W1b1 W2b2


















100

075

050

025

000-4 -2 0

Sigmoid

2 4

1

05

0

ndash05

ndash1-4 -2 0

Tanh

2 4









Tourismfacilities

usesaturation

080- 095 070- 1 00 0 00- 010 0 00- 010 000- 005 070- 100 0 45- 1 00 0 00- 010 000- 005 0 00- 010 Securitylevel

070- 080 060- 0 70 0 10- 015 0 10- 015 005- 006 060- 0 70 0 35- 0 45 0 10- 020 005- 010 0 10- 020 A goodlevel

060- 070 050- 0 60 0 15- 020 0 15- 020 006- 008 050- 0 60 0 25- 0 35 0 20- 030 010- 020 0 20- 030 qualified


Politicalstability

Frequencyof

occurrenceof


Frequencyof

earthquakeand

geologicaldisasters



Trafficsafety

Theconsumer

price indexrose

Thefrequency

ofoutbreaks

ofepidemicdiseases


guest

Alarmoutput

indicatingsignal


06

05

04

03

02

01

00 200 400



Root

mea

n sq

uare

erro

r

600 800









218161412

Root

mea

n sq

uare

erro

r

108060402

00 200 400




09

08

07

06

05

04

Loss

func

tion

valu

e

03

02

01

00 5 10 15

Number of training


20 25 30


15

1

05

0

Erro

r dev

iatio

n

ndash05

ndash1230 250 270 290 310 330

e serial number


350 370 390 410 430 450 470 490 510





3 Conclusion



Warninginstructions




Police line on

The p line

Time






Data Availability




Acknowledgments


References





































h f W1T

x + b1

1113872 1113873 (1)

y f W2T

x + b2

1113872 1113873 (2)


f 1

1 + exp(minusx) (3)






⎧⎪⎪⎪⎪⎪⎪⎨

⎪⎪⎪⎪⎪⎪⎩

(4)


Δθ(n + 1) θ(n + 1) minus θ(n)1113896 (5)



ht f W1T

xt + WhT

htminus1 + b1

1113872 1113873 (6)

yt f W2T

ht + b2

1113872 1113873 (7)


tanhx sinhx

cosh x

ex

minus eminusx

ex

+ eminusx (8)


Lt

minusytlog o

t (9)


L x1 x

t1113966 1113967 y

1 y

t1113966 11139671113872 1113873 1113944 L

t (10)








LSTM neural network


Warning range



Safety zone




move incircles

move incircles






Lin

Lhidden

Lout

x

h

y


h tndash2

y tndash2

h tndash1

y tndash1

h

y

Delayspread

W1b1

W2b2

WhW1b1

W2b2

W1b1

W2b2

W1b1

W2b2

WhWhWh


X +

X X

tabhδ δ δ

tanh

xt

ht

A A

xtndash1 xt+1

htndash1 ht+1


Lin Lhidden Lout

x1

x3

h3

y1

x2

h2

h4

y3

h1

y2

W1b1 W2b2


















100

075

050

025

000-4 -2 0

Sigmoid

2 4

1

05

0

ndash05

ndash1-4 -2 0

Tanh

2 4









Tourismfacilities

usesaturation

080- 095 070- 1 00 0 00- 010 0 00- 010 000- 005 070- 100 0 45- 1 00 0 00- 010 000- 005 0 00- 010 Securitylevel

070- 080 060- 0 70 0 10- 015 0 10- 015 005- 006 060- 0 70 0 35- 0 45 0 10- 020 005- 010 0 10- 020 A goodlevel

060- 070 050- 0 60 0 15- 020 0 15- 020 006- 008 050- 0 60 0 25- 0 35 0 20- 030 010- 020 0 20- 030 qualified


Politicalstability

Frequencyof

occurrenceof


Frequencyof

earthquakeand

geologicaldisasters



Trafficsafety

Theconsumer

price indexrose

Thefrequency

ofoutbreaks

ofepidemicdiseases


guest

Alarmoutput

indicatingsignal


06

05

04

03

02

01

00 200 400



Root

mea

n sq

uare

erro

r

600 800









218161412

Root

mea

n sq

uare

erro

r

108060402

00 200 400




09

08

07

06

05

04

Loss

func

tion

valu

e

03

02

01

00 5 10 15

Number of training


20 25 30


15

1

05

0

Erro

r dev

iatio

n

ndash05

ndash1230 250 270 290 310 330

e serial number


350 370 390 410 430 450 470 490 510





3 Conclusion



Warninginstructions




Police line on

The p line

Time






Data Availability




Acknowledgments


References




































LSTM neural network


Warning range



Safety zone




move incircles

move incircles






Lin

Lhidden

Lout

x

h

y


h tndash2

y tndash2

h tndash1

y tndash1

h

y

Delayspread

W1b1

W2b2

WhW1b1

W2b2

W1b1

W2b2

W1b1

W2b2

WhWhWh


X +

X X

tabhδ δ δ

tanh

xt

ht

A A

xtndash1 xt+1

htndash1 ht+1


Lin Lhidden Lout

x1

x3

h3

y1

x2

h2

h4

y3

h1

y2

W1b1 W2b2


















100

075

050

025

000-4 -2 0

Sigmoid

2 4

1

05

0

ndash05

ndash1-4 -2 0

Tanh

2 4









Tourismfacilities

usesaturation

080- 095 070- 1 00 0 00- 010 0 00- 010 000- 005 070- 100 0 45- 1 00 0 00- 010 000- 005 0 00- 010 Securitylevel

070- 080 060- 0 70 0 10- 015 0 10- 015 005- 006 060- 0 70 0 35- 0 45 0 10- 020 005- 010 0 10- 020 A goodlevel

060- 070 050- 0 60 0 15- 020 0 15- 020 006- 008 050- 0 60 0 25- 0 35 0 20- 030 010- 020 0 20- 030 qualified


Politicalstability

Frequencyof

occurrenceof


Frequencyof

earthquakeand

geologicaldisasters



Trafficsafety

Theconsumer

price indexrose

Thefrequency

ofoutbreaks

ofepidemicdiseases


guest

Alarmoutput

indicatingsignal


06

05

04

03

02

01

00 200 400



Root

mea

n sq

uare

erro

r

600 800









218161412

Root

mea

n sq

uare

erro

r

108060402

00 200 400




09

08

07

06

05

04

Loss

func

tion

valu

e

03

02

01

00 5 10 15

Number of training


20 25 30


15

1

05

0

Erro

r dev

iatio

n

ndash05

ndash1230 250 270 290 310 330

e serial number


350 370 390 410 430 450 470 490 510





3 Conclusion



Warninginstructions




Police line on

The p line

Time






Data Availability




Acknowledgments


References































Lin

Lhidden

Lout

x

h

y


h tndash2

y tndash2

h tndash1

y tndash1

h

y

Delayspread

W1b1

W2b2

WhW1b1

W2b2

W1b1

W2b2

W1b1

W2b2

WhWhWh


X +

X X

tabhδ δ δ

tanh

xt

ht

A A

xtndash1 xt+1

htndash1 ht+1


Lin Lhidden Lout

x1

x3

h3

y1

x2

h2

h4

y3

h1

y2

W1b1 W2b2


















100

075

050

025

000-4 -2 0

Sigmoid

2 4

1

05

0

ndash05

ndash1-4 -2 0

Tanh

2 4









Tourismfacilities

usesaturation

080- 095 070- 1 00 0 00- 010 0 00- 010 000- 005 070- 100 0 45- 1 00 0 00- 010 000- 005 0 00- 010 Securitylevel

070- 080 060- 0 70 0 10- 015 0 10- 015 005- 006 060- 0 70 0 35- 0 45 0 10- 020 005- 010 0 10- 020 A goodlevel

060- 070 050- 0 60 0 15- 020 0 15- 020 006- 008 050- 0 60 0 25- 0 35 0 20- 030 010- 020 0 20- 030 qualified


Politicalstability

Frequencyof

occurrenceof


Frequencyof

earthquakeand

geologicaldisasters



Trafficsafety

Theconsumer

price indexrose

Thefrequency

ofoutbreaks

ofepidemicdiseases


guest

Alarmoutput

indicatingsignal


06

05

04

03

02

01

00 200 400



Root

mea

n sq

uare

erro

r

600 800









218161412

Root

mea

n sq

uare

erro

r

108060402

00 200 400




09

08

07

06

05

04

Loss

func

tion

valu

e

03

02

01

00 5 10 15

Number of training


20 25 30


15

1

05

0

Erro

r dev

iatio

n

ndash05

ndash1230 250 270 290 310 330

e serial number


350 370 390 410 430 450 470 490 510





3 Conclusion



Warninginstructions




Police line on

The p line

Time






Data Availability




Acknowledgments


References














































100

075

050

025

000-4 -2 0

Sigmoid

2 4

1

05

0

ndash05

ndash1-4 -2 0

Tanh

2 4









Tourismfacilities

usesaturation

080- 095 070- 1 00 0 00- 010 0 00- 010 000- 005 070- 100 0 45- 1 00 0 00- 010 000- 005 0 00- 010 Securitylevel

070- 080 060- 0 70 0 10- 015 0 10- 015 005- 006 060- 0 70 0 35- 0 45 0 10- 020 005- 010 0 10- 020 A goodlevel

060- 070 050- 0 60 0 15- 020 0 15- 020 006- 008 050- 0 60 0 25- 0 35 0 20- 030 010- 020 0 20- 030 qualified


Politicalstability

Frequencyof

occurrenceof


Frequencyof

earthquakeand

geologicaldisasters



Trafficsafety

Theconsumer

price indexrose

Thefrequency

ofoutbreaks

ofepidemicdiseases


guest

Alarmoutput

indicatingsignal


06

05

04

03

02

01

00 200 400



Root

mea

n sq

uare

erro

r

600 800









218161412

Root

mea

n sq

uare

erro

r

108060402

00 200 400




09

08

07

06

05

04

Loss

func

tion

valu

e

03

02

01

00 5 10 15

Number of training


20 25 30


15

1

05

0

Erro

r dev

iatio

n

ndash05

ndash1230 250 270 290 310 330

e serial number


350 370 390 410 430 450 470 490 510





3 Conclusion



Warninginstructions




Police line on

The p line

Time






Data Availability




Acknowledgments


References





































Tourismfacilities

usesaturation

080- 095 070- 1 00 0 00- 010 0 00- 010 000- 005 070- 100 0 45- 1 00 0 00- 010 000- 005 0 00- 010 Securitylevel

070- 080 060- 0 70 0 10- 015 0 10- 015 005- 006 060- 0 70 0 35- 0 45 0 10- 020 005- 010 0 10- 020 A goodlevel

060- 070 050- 0 60 0 15- 020 0 15- 020 006- 008 050- 0 60 0 25- 0 35 0 20- 030 010- 020 0 20- 030 qualified


Politicalstability

Frequencyof

occurrenceof


Frequencyof

earthquakeand

geologicaldisasters



Trafficsafety

Theconsumer

price indexrose

Thefrequency

ofoutbreaks

ofepidemicdiseases


guest

Alarmoutput

indicatingsignal


06

05

04

03

02

01

00 200 400



Root

mea

n sq

uare

erro

r

600 800









218161412

Root

mea

n sq

uare

erro

r

108060402

00 200 400




09

08

07

06

05

04

Loss

func

tion

valu

e

03

02

01

00 5 10 15

Number of training


20 25 30


15

1

05

0

Erro

r dev

iatio

n

ndash05

ndash1230 250 270 290 310 330

e serial number


350 370 390 410 430 450 470 490 510





3 Conclusion



Warninginstructions




Police line on

The p line

Time






Data Availability




Acknowledgments


References





































218161412

Root

mea

n sq

uare

erro

r

108060402

00 200 400




09

08

07

06

05

04

Loss

func

tion

valu

e

03

02

01

00 5 10 15

Number of training


20 25 30


15

1

05

0

Erro

r dev

iatio

n

ndash05

ndash1230 250 270 290 310 330

e serial number


350 370 390 410 430 450 470 490 510





3 Conclusion



Warninginstructions




Police line on

The p line

Time






Data Availability




Acknowledgments


References

































3 Conclusion



Warninginstructions




Police line on

The p line

Time






Data Availability




Acknowledgments


References
































Data Availability




Acknowledgments


References































Optimization Algorithm of Tourism Security Early Warning ...

Documents

Transcript of Optimization Algorithm of Tourism Security Early Warning ...