A scenic spot reception number pre-judgment method and device based on a big data platform

Abstract

The application discloses a scenic spot reception number prediction method and device based on a big data platform, real-time current tourist quantity, train ticket ordering quantity of related scenic spots and search times of a certain scenic spot in a map APP are acquired by using the big data platform, a soft measurement modeling method of just-in-time learning (JITL) is adopted, the nearest neighbor matching data is found from historical data by using a previous section of data at a current time to perform local modeling, real-time tourists of the scenic spot are adaptively updated and predicted, and the scenic spot tourist saturation prediction and judgment positively affect decision-making.

Description

Technical Field

[0001] This invention relates to the field of smart tourism technology, and in particular to a method and device for predicting the number of visitors to a scenic spot based on a big data platform. Background Technology

[0002] In today's rapidly developing society, with advanced technology and information flow, communication between people is becoming increasingly close, and life is becoming more and more convenient. Big data is one of the products of this high-tech era. Data mining can be divided into online and offline methods. Offline technologies include various techniques such as facial recognition, fingerprint recognition, and WiFi probes that collect biometric features from mobile devices.

[0003] Online technology can use web crawlers to obtain real-time information on train ticket bookings for a particular attraction, as well as real-time search volume for that attraction on Baidu Maps or Gaode Maps. By integrating this data along a time dimension and fully exploring the correlations between the data, it is possible to predict and provide early warnings about the number of visitors to a particular attraction.

[0004] Since the number of tourists and the number of tourists arriving at a scenic spot are changing in real time, it is difficult to adaptively model the number of tourists in the next moment using a global model, thus limiting its early warning function. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a method and device for predicting the number of visitors to a scenic spot based on a big data platform, so as to more accurately predict the number of visitors in the next moment.

[0006] In a first aspect, the present invention provides a method for predicting the number of visitors to a scenic area based on a big data platform, including:

[0007] Step 10: Based on the big data platform, obtain the real-time reservation data of train tickets for designated scenic spots and the real-time search count of designated scenic spots in the electronic map, and integrate the train ticket reservation data and real-time search count by time dimension to obtain a historical dataset.

[0008] Step 20: Obtain the current reservation data of train tickets for the designated scenic spot and the real-time search count of the designated scenic spot in the electronic map as the test dataset. Use the cloud model to measure the similarity between all samples in the historical dataset and the test dataset, and take the top k samples as the nearest neighbor data of the test dataset.

[0009] Step 30: Using the number of tourists in the specified scenic area at the next moment as the dependent variable and the nearest neighbor data as the independent variable, introduce Lagrange multipliers to calculate the weight vectors of the independent and dependent variables. Then, extract the factors of the latent variables of the independent and dependent variables and the regression coefficients between them. Next, calculate the residual space of the independent and dependent variables, train the partial least squares model to minimize the residual space, and finally establish a linear regression equation between the residual spaces and perform inverse standardization to obtain the local model corresponding to the test dataset.

[0010] Step 40: Input the current reservation data of train tickets for the designated scenic spot and the real-time search count of the designated scenic spot in the electronic map into the local model to obtain the estimated number of tourists in the designated scenic spot at the next moment.

[0011] Furthermore, in step 20, the similarity measurement between all samples in the historical dataset and the test dataset is performed using a cloud model, specifically including:

[0012] Step 21: Calculate the reservation data for train tickets to the designated scenic spot and the quantitative value x of the real-time search volume for the designated scenic spot on the electronic map. i sample mean First-order sample absolute central moments Sample variance

[0013] Step 22: Calculate the expected value of the cloud distribution from the sample mean in Step 21:

[0014] Step 23: Calculate the entropy of the cloud distribution using the sample mean from Step 21 and the expected value of the cloud distribution from Step 22.

[0015] Step 24: The hyperentropy of the cloud distribution can be calculated from the sample variance in Step 21 and the entropy of the cloud distribution in Step 23.

[0016] Step 25: Calculate the cloud model digital features of the historical dataset and the test dataset using the methods from Steps 21 to 24. Then, calculate the similarity between the cloud model digital features of the historical dataset and the cloud model digital features of the test dataset, using the following formula:

[0017]

[0018] Where α=(Ex α En α He α ) represents the digital features of the cloud model in the historical dataset, β = (Ex β En β Heβ ) represents the digital features of the cloud model in the test dataset.

[0019] Furthermore, step 30 specifically includes:

[0020] Step 31: Using the reservation data of train tickets for the designated scenic spot and the real-time search volume of the designated scenic spot on the electronic map as the sample input quantity p, and the number of visitors to the scenic spot as the sample output quantity q, there are p input quantities {x1, x2, ..., x}. p} and q output quantities {y1, y2, ..., y q Let} be the sample, and let the input data matrix X∈R be the sample. n×p Output data matrix Y∈R n×q Let the first principal component axis vectors of X and Y be w1(p×1) and c1(q×1), respectively, and both be unit vectors. Then w1 and c1 represent the first principal components t1 and u1 of X and Y, where t1 = X*w1 and u1 = Y*c1.

[0021] Step 32: Maximize the correlation between t1 and u1 using partial least squares method, as shown in the following formula:

[0022] Maximize<Xw1,Yc1> ,Subject to:||w1||=1,||c1||=1

[0023] Among them, Maximize<Xw1,Yc1> This represents maximizing the correlation between t1 and u1. t1 is the first component of X and the first axis of X. It is a unit vector, i.e., ||w1|| = 1. Similarly, we get ||c1|| = 1.

[0024] Step 33: Use the Lagrange multiplier method to find the weight vector w of the independent variable. k and the weight vector c of the dependent variable k , where w k is u' k v k v' k u k The eigenvector corresponding to the largest eigenvalue, c k v' k u k u' k v k The eigenvector corresponding to the largest eigenvalue; its independent variable is the latent variable t. k and the latent variable s of the dependent variable k The calculation is as follows:

[0025] t k =u k-1 w k

[0026] s k =v k-1 c k

[0027] Step 34: Extract the factor loadings p of the latent independent and dependent variables. k and q k and the regression coefficient b between the two k :

[0028] p' k =(t' k t k ) -1 t' k u k-1

[0029] q' k =(t' k t k ) -1 t' k v k-1

[0030] b' k =(t' k t k ) -1 t' k s k

[0031] Step 35: Calculate the residual space u of the independent variable x and the dependent variable y. k v k :

[0032] u k =u k-1 -t k p' k

[0033] v k =v k-1 -t k q' k

[0034] Step 36: Determine the residual space u k and v k Has the minimum value been reached? If not, return to step 32. If the minimum value has been reached, establish u. k and v k The linear regression equation between them:

[0035] v=uβ

[0036]

[0037] Step 36: Through inverse standardization, obtain the regression equations for the dependent variable Y and the independent variable X:

[0038] Y = Xβ.

[0039] Secondly, the present invention provides a device for predicting the number of visitors to a scenic area based on a big data platform, comprising: a data preprocessing module, a cloud model similarity measurement module, a local model construction module, and a prediction output module;

[0040] The data preprocessing module is used to acquire real-time reservation data of train tickets for designated scenic spots and real-time search counts of designated scenic spots in electronic maps based on a big data platform, and integrate the reservation data of train tickets and real-time search counts according to the time dimension to obtain a historical dataset.

[0041] The cloud model similarity measurement module is used to obtain the current reservation data of train tickets for the specified scenic spot and the real-time search count of the specified scenic spot in the electronic map as the test dataset. The cloud model measures the similarity between all samples in the historical dataset and the test dataset, and extracts the top k samples as the nearest neighbor data of the test dataset.

[0042] The local model construction module is used to take the number of tourists in the specified scenic area at the next moment as the dependent variable and the nearest neighbor data as the independent variable. It introduces Lagrange multipliers to calculate the weight vectors of the independent and dependent variables, then extracts the factors of the latent variables of the independent and dependent variables and the regression coefficients between them, then calculates the residual space of the independent and dependent variables, trains the partial least squares model to minimize the residual space, and finally establishes a linear regression equation between the residual spaces and performs inverse standardization to obtain the local model corresponding to the test dataset.

[0043] The prediction output module is used to input the current reservation data of train tickets for the specified scenic spot and the real-time search count of the specified scenic spot in the electronic map into the local model to obtain the estimated number of tourists in the specified scenic spot at the next moment.

[0044] Furthermore, the cloud model similarity measurement module measures the similarity between all samples in the historical dataset and the test dataset using the cloud model, specifically including the following steps:

[0045] Step 21: Calculate the reservation data for train tickets to the designated scenic spot and the quantitative value x of the real-time search volume for the designated scenic spot on the electronic map. i sample mean First-order sample absolute central moments Sample variance

[0046] Step 22: Calculate the expected value of the cloud distribution from the sample mean in Step 21:

[0047] Step 23: Calculate the entropy of the cloud distribution using the sample mean from Step 21 and the expected value of the cloud distribution from Step 22.

[0048] Step 24: The hyperentropy of the cloud distribution can be calculated from the sample variance in Step 21 and the entropy of the cloud distribution in Step 23.

[0049] Step 25: Calculate the cloud model digital features of the historical dataset and the test dataset using the methods from Steps 21 to 24. Then, calculate the similarity between the cloud model digital features of the historical dataset and the cloud model digital features of the test dataset, using the following formula:

[0050]

[0051] Where α=(Ex α En α He α ) represents the digital features of the cloud model in the historical dataset, β = (Ex β En β He β ) represents the digital features of the cloud model in the test dataset.

[0052] Furthermore, the local model construction module is specifically used to perform the following steps:

[0053] Step 31: Using the reservation data of train tickets for the designated scenic spot and the real-time search volume of the designated scenic spot on the electronic map as the sample input quantity p, and the number of visitors to the scenic spot as the sample output quantity q, there are p input quantities {x1, x2, ..., x}. p} and q output quantities {y1, y2, ..., y q Let} be the sample, and let the input data matrix X∈R be the sample. n×p Output data matrix Y∈R n×q Let the first principal component axis vectors of X and Y be w1(p×1) and c1(q×1), respectively, and both be unit vectors. Then w1 and c1 represent the first principal components t1 and u1 of X and Y, where t1 = X*w1 and u1 = Y*c1.

[0054] Step 32: Maximize the correlation between t1 and u1 using partial least squares method, as shown in the following formula:

[0055] Maximize<Xw1,Yc1> ,Subject to:||w1||=1,||c1||=1

[0056] Among them, Maximize<Xw1,Yc1> This represents maximizing the correlation between t1 and u1. t1 is the first component of X and the first axis of X. It is a unit vector, i.e., ||w1|| = 1. Similarly, we get ||c1|| = 1.

[0057] Step 33: Use the Lagrange multiplier method to find the weight vector w of the independent variable. k and the weight vector c of the dependent variable k , where w k is u' k v k v' k u k The eigenvector corresponding to the largest eigenvalue, c k v' k u k u' k v k The eigenvector corresponding to the largest eigenvalue; its independent variable is the latent variable t. k and the latent variable s of the dependent variable k The calculation is as follows:

[0058] t k =u k-1 w k

[0059] s k =v k-1 c k

[0060] Step 34: Extract the factor loadings p of the latent independent and dependent variables. k and q k and the regression coefficient b between the two k :

[0061] p' k =(t' k t k ) -1 t' k u k-1

[0062] q' k =(t' k t k ) -1 t' k v k-1

[0063] b' k =(t' k t k ) -1 t' k s k

[0064] Step 35: Calculate the residual space u of the independent variable x and the dependent variable y. k v k :

[0065] u k =u k-1 -t k p' k

[0066] v k =v k-1 -t k q' k

[0067] Step 36: Determine the residual space u k and v k Has the minimum value been reached? If not, return to step 32. If the minimum value has been reached, establish u. k and v k The linear regression equation between them:

[0068] v=uβ

[0069]

[0070] Step 36: Through inverse standardization, obtain the regression equations for the dependent variable Y and the independent variable X:

[0071] Y = Xβ.

[0072] One or more technical solutions provided in the embodiments of the present invention have at least the following technical effects or advantages:

[0073] The Just-In-Time Learning (JITL) soft measurement modeling method is adopted. It uses the previous data of the current moment to find the nearest neighbor matching data from the historical data to perform local modeling, and adaptively updates and predicts the real-time tourists in the scenic area.

[0074] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and in order to make the above and other objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description

[0075] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0076] Figure 1 This is a model framework diagram of the method in Embodiment 1 of the present invention;

[0077] Figure 2 This is a schematic diagram of the device in Embodiment 2 of the present invention. Detailed Implementation

[0078] This application provides a method and apparatus for predicting the number of visitors to a scenic area based on a big data platform, enabling more accurate prediction of the number of visitors at the next moment.

[0079] The overall concept of the technical solution in this application is as follows:

[0080] Based on a big data platform, real-time data is acquired from WiFi probes, including train ticket booking information for a specific tourist attraction, and real-time search volume for that attraction on Baidu Maps or Gaode Maps. These three types of data are integrated along a time dimension to mine correlations, train a model, and provide early warning and prediction of visitor numbers for a particular attraction. Based on this, a Just-In-Time Learning (JITL) soft measurement modeling method is employed. This method uses previous data from the current moment to find the nearest neighbor matching data from historical data for local modeling, enabling adaptive updates and predictions of real-time visitor numbers at the scenic area. The similarity measurement method is a core component of JITL. Traditional JITL uses various distance metrics, such as Euclidean distance and Manhattan distance, which cannot perfectly capture changes in visitor volume. Therefore, using a cloud model for similarity measurement can improve the quality of nearest neighbor sample selection.

[0081] Example 1

[0082] This embodiment provides a method for predicting the number of visitors to a scenic area based on a big data platform, including:

[0083] Step 10: Based on the big data platform, obtain the real-time reservation data of train tickets for designated scenic spots and the real-time search count of designated scenic spots in the electronic map, and integrate the train ticket reservation data and real-time search count by time dimension to obtain a historical dataset.

[0084] Step 20: Obtain the current reservation data of train tickets for the designated scenic spot and the real-time search count of the designated scenic spot in the electronic map as the test dataset. Use the cloud model to measure the similarity between all samples in the historical dataset and the test dataset, and take the top k samples as the nearest neighbor data of the test dataset (the value of k can be set as needed).

[0085] The cloud model, through a specific structural algorithm, realizes a model for the uncertainty transformation between qualitative concepts and their quantitative representations. Taking the number of visitors to a certain scenic spot as the quantitative universe of discourse U, and the number of train ticket purchases and map searches as cloud drops x, the distribution of x on the universe of discourse U is called the cloud model, and each x is called a cloud drop. The inverse cloud algorithm takes the quantitative values ​​as input and outputs three numerical features of the qualitative concept cloud model: expectation Ex, entropy En, and hyperentropy He. The similarity between the preprocessed test data and all samples in the historical dataset is obtained through a cloud model similarity metric, and the top k samples are taken as the nearest neighbors of the current test data.

[0086] (I) Introduction to the Cloud Model

[0087] The cloud model, developed by Professor Li Deyi based on probability and statistics theory and fuzzy mathematics theory, utilizes a specific structural algorithm to achieve the uncertainty transformation between qualitative concepts and their quantitative representations. Since its inception and application, the cloud model has been defined in two ways. The initial definition proposed by Professor Li Deyi is as follows:

[0088] Definition 1: Let U be the quantitative universe of discourse (specifically, the number of visitors to a certain scenic spot in this embodiment), and C be a qualitative concept on U (a fuzzy set of the number of visitors to a scenic spot in this embodiment). If Each has a random number μ with a stable tendency. c (x), is called the membership degree of x to C, μ c The distribution of (x) on U is called a membership cloud, and (x,μ) is called a cloud droplet (in this embodiment, it is the number of train ticket purchases or search results for a certain attraction in a map app).

[0089] With the gradual development and application of cloud model research, Professor Li Deyi's team proposed an alternative definition for the convenience of theoretical discussion and applied research, and gradually replaced the previous one.

[0090] Definition 2: Let U be the quantitative universe of discourse, C be a qualitative concept on U, x be a random variable X defined on U and realized once, and let μ be the membership degree of x to C. c (x)[0,1] is a random variable with a stable tendency.

[0091]

[0092] The distribution of X on the universe of discourse U is called a cloud, and x is called a cloud droplet. For any c∈U, the mapping of x to the interval [0,1] is a one-to-many transformation. The certainty of x with respect to C is not a fixed value, but a probability distribution. A cloud is composed of cloud dropslets (the number of train ticket bookings or search results for a certain attraction in a map app). Since the number of train ticket bookings and attraction searches is unordered, the number of cloud dropslets is also unordered. A cloud droplet is a quantitative realization of a qualitative concept. The more cloud dropslets there are, the better they reflect the overall characteristics of this qualitative concept. The higher the probability of cloud drops appearing and the higher the certainty of cloud drops, the greater their contribution to the concept. That is, the more train ticket bookings and attraction searches, the better they reflect its overall characteristics. The overall characteristics of the cloud model are usually represented by the following three numerical features:

[0093] 1) Expectation (Ex): The expected value of the number of train ticket bookings and app searches for scenic spots in the domain space, which is the central point of the concept in the domain. The value in the domain corresponding to the centroid of the area covered by the cloud is the point that best represents the qualitative concept. In the cloud map, this is reflected as a point with a membership degree of 1.

[0094] 2) Entropy (En): A measure of randomness and fuzziness in qualitative concepts. Randomness is measured by the discreteness of cloud droplets; while fuzziness is measured by the range of values ​​that cloud droplets can take in the universe of discourse. It is reflected in the cloud shape as the "span" of the cloud; the greater the entropy, the larger the "span" of the cloud. Using the same numerical characteristic to reflect randomness and fuzziness bridges the gap between the two. It reflects the range of cloud droplets that are conceptually acceptable in the universe of discourse.

[0095] 3) Hyperentropy (He): A measure of the uncertainty of entropy, i.e., the entropy of entropy. It is determined by the randomness and fuzziness of entropy. Reflected in cloud maps, it manifests as the dispersion of cloud droplets and the "thickness" of the cloud. Generally, the greater the hyperentropy, the "thicker" the cloud, and the thickness varies from place to place. Clouds closer to and farther from the concept center are thinner, clouds in between are thicker, and the cloud thickness is greatest where the membership degree has the largest variance.

[0096] In both natural and social sciences, the most prevalent and widely used probability distribution function is the normal distribution, commonly characterized by its expectation and variance. Similarly, the normal membership function, widely used in fuzzy sets, is expressed as μC(x)=exp(-(xa)). 2 / 2b 2 The normal cloud model is a transformation model developed and evolved from these two foundations. This model combines the randomness and fuzziness existing in the objective world, forming a mutual mapping between qualitative and quantitative aspects. Its definition is as follows:

[0097] Definition 3: Given a universe of discourse U, the distribution of cloud droplets on U generated by the cloud model C(Ex, En, He) satisfies:

[0098] x~N(Ex,En′ 2 (1.2)

[0099] The distribution of cloud droplets generated by the cloud model for the number of visitors to a certain scenic spot follows a normal distribution, while the number of train ticket purchases and app searches for the scenic spot are normal clouds over U. Where En'~N(En,He) 2 ),N(En,He 2 N(Ex,En′) represents a normal distribution with expected value En and standard deviation He. 2 Let represent a normal distribution with expectation Ex and standard deviation En′. The membership degree of x to concept C is μ(x) = exp(-(x-Ex)). 2 / 2), the distribution of X on the universe of discourse U is called a normal cloud.

[0100] The inverse normal cloud generator completes the transformation from quantitative values ​​to qualitative concepts. Through statistical analysis or other effective techniques, it efficiently converts a certain number of quantitative values ​​into qualitative concepts expressed by the three numerical characteristics of a cloud model. Its inverse cloud algorithm is shown below:

[0101]

[0102] (II) Calculation of Cloud Model Similarity

[0103] Given a vector α = (Ex) consisting of two cloud digital features α En α He α ), β = (Ex β En β He β Let α and β be the cloud formed by two cloud droplets: the booking status of train tickets for a certain tourist attraction and the real-time search volume for that attraction on Baidu Maps or Gaode Maps. The similarity between these two clouds α and β is expressed by the cosine angle between them:

[0104]

[0105] Where α=(Ex α En α He α ) represents the digital features of the cloud model in the historical dataset, β = (Ex β En β He β ) represents the digital features of the cloud model in the test dataset.

[0106] Step 30: Using the number of tourists in the specified scenic area at the next moment as the dependent variable and the nearest neighbor data as the independent variable, introduce Lagrange multipliers to calculate the weight vectors of the independent and dependent variables. Then, extract the factors of the latent variables of the independent and dependent variables and the regression coefficients between them. Next, calculate the residual space of the independent and dependent variables, train the partial least squares model to minimize the residual space, and finally establish a linear regression equation between the residual spaces and perform inverse standardization to obtain the local model corresponding to the test dataset.

[0107] Partial least squares (PLS) is a regression modeling method for multiple dependent variables on multiple independent variables. It performs orthogonal decomposition on the input and output vectors simultaneously to obtain as much information as possible and can remove the interactions between data. Its basic idea is to orthogonally project a high-dimensional space onto a low-dimensional feature space, thereby overcoming the collinearity problem in least squares regression through orthogonal feature projection.

[0108] Suppose there are p input variables {x1, x2, ..., xn} p} and q output quantities {y1, y2, ..., y q Let} be the sample, and let the input data matrix X∈R be the sample. n×p Output data matrix Y∈R n×q Let the first principal component axis vectors of X and Y be w1 (p×1) and c1 (q×1), both of which are unit vectors. Then, w1 and c1 can represent the first principal components t1 and u1 of X and Y, where t1 = X*w1 and u1 = Y*c1. The basic idea of ​​partial least squares is to maximize the correlation between t1 and u1, which can be mathematically represented as:

[0109] Maximize<Xw1,Yc1> ,Subject to:||w1||=1,||c1||=1 (1.4)

[0110] The method of introducing Lagrange multipliers is used to find the weight vector w of the independent variable. k and the weight vector c of the dependent variable k , where w k is u' k v k v' k u k The eigenvector corresponding to the largest eigenvalue, c k v' k u k u' k v k The eigenvector corresponding to the largest eigenvalue. Its independent variable is the latent variable t. k and the latent variable s of the dependent variable k The calculation is as follows:

[0111]

[0112] Further calculations were performed to extract the factor loadings p of the latent independent and dependent variables. k and q k and the regression coefficient b between the two k :

[0113]

[0114] Calculate the residual space u of the independent variable x and the dependent variable y. k v k :

[0115]

[0116] Determine the residual space u k and v k If the minimum value has not been reached, repeat formula 1.4 until the residual space reaches its minimum, and then establish u. k and v k The linear regression equation between them:

[0117]

[0118] Through inverse standardization, the regression equations for the dependent variable Y and the independent variable X are finally obtained:

[0119] Y = Xβ (1.9)

[0120] Step 40: Input the current reservation data of train tickets for the designated scenic spot and the real-time search count of the designated scenic spot in the electronic map into the local model to obtain the estimated number of tourists in the designated scenic spot at the next moment.

[0121] The final regression equation obtained is the relationship between the dependent variable (the number of visitors to the attraction at the next moment) and the independent variables (train ticket booking status and the number of searches for the attraction on the map app). This equation can more accurately predict the number of visitors to the scenic spot at the next moment and play a predictive and early warning role.

[0122] Please refer to Figure 1 Within the framework of real-time learning, a cloud model is used to select neighboring data similar to the previous data at the current time, and a partial least squares model is used to predict the tourist data at the next time step.

[0123] The specific implementation steps are as follows:

[0124] (1) Data preprocessing: Based on the big data platform, the data received by the WiFi probe is obtained in real time, and the booking status of train tickets for a certain scenic spot and the real-time search volume of a certain scenic spot in Baidu Map or Gaode Map are obtained in real time through crawling. These data are integrated together according to the time dimension to become a three-dimensional training set x.

[0125] (2) Cloud Model Similarity Measurement: Using a cloud model, a model for the uncertainty transformation between qualitative concepts and their quantitative representations is realized through a specific structural algorithm. The number of visitors to a scenic spot is taken as the quantitative universe of discourse U, and the number of train ticket purchases and map searches are taken as cloud droplets x. The distribution of x in the universe of discourse U is called the cloud model, and each x is called a cloud droplet. The inverse cloud algorithm is used to input quantitative values ​​and output three numerical features reflecting the qualitative concept cloud model: expectation Ex, entropy En, and hyperentropy He. The similarity between the preprocessed test data and all samples in the historical dataset is obtained through cloud model similarity measurement, and the top k samples are taken as the nearest neighbors of the current test data.

[0126] (3) Training the Partial Least Squares (PLS) Model: Partial least squares orthogonally projects a high-dimensional space onto a low-dimensional feature space, overcoming the collinearity problem in least squares regression through orthogonal feature projection. The basic idea of ​​partial least squares is to maximize the correlation between t1 and u1. The weight vectors of the independent and dependent variables are obtained by introducing Lagrange multipliers. Further calculations are performed to extract the factors of the latent variables of the independent and dependent variables, as well as the regression coefficients between them. The residual spaces of the independent and dependent variables are calculated, and a linear regression equation between the residual spaces is established and inversely standardized to obtain the local model corresponding to the current test data.

[0127] (4) Predicted output: Input the current test sample into the established PLS model to obtain the material level estimate.

[0128] By utilizing big data platforms to obtain real-time data on current tourist volume, train ticket bookings for relevant attractions, and search volume for specific attractions on map apps, the number of tourists at the next moment can be predicted, playing a positive role in predicting and judging the saturation level of tourist attractions.

[0129] Based on the same inventive concept, this application also provides an apparatus corresponding to the method in Embodiment 1, as detailed in Embodiment 2.

[0130] Example 2

[0131] This embodiment provides a device for predicting the number of visitors to a scenic area based on a big data platform, such as... Figure 2 As shown, it includes: a data preprocessing module, a cloud model similarity measurement module, a local model construction module, and a prediction output module;

[0132] The data preprocessing module is used to acquire real-time reservation data of train tickets for designated scenic spots and real-time search counts of designated scenic spots in electronic maps based on a big data platform, and integrate the reservation data of train tickets and real-time search counts according to the time dimension to obtain a historical dataset.

[0133] The cloud model similarity measurement module is used to obtain the current reservation data of train tickets for the specified scenic spot and the real-time search count of the specified scenic spot in the electronic map as the test dataset. The cloud model measures the similarity between all samples in the historical dataset and the test dataset, and extracts the top k samples as the nearest neighbor data of the test dataset.

[0134] The local model construction module is used to take the number of tourists in the specified scenic area at the next moment as the dependent variable and the nearest neighbor data as the independent variable. It introduces Lagrange multipliers to calculate the weight vectors of the independent and dependent variables, then extracts the factors of the latent variables of the independent and dependent variables and the regression coefficients between them, then calculates the residual space of the independent and dependent variables, trains the partial least squares model to minimize the residual space, and finally establishes a linear regression equation between the residual spaces and performs inverse standardization to obtain the local model corresponding to the test dataset.

[0135] The prediction output module is used to input the current reservation data of train tickets for the specified scenic spot and the real-time search count of the specified scenic spot in the electronic map into the local model to obtain the estimated number of tourists in the specified scenic spot at the next moment.

[0136] Furthermore, the cloud model similarity measurement module measures the similarity between all samples in the historical dataset and the test dataset using the cloud model, specifically including the following steps:

[0137] Step 21: Calculate the reservation data for train tickets to the designated scenic spot and the quantitative value x of the real-time search volume for the designated scenic spot on the electronic map. i sample mean First-order sample absolute central moments Sample variance

[0138] Step 22: Calculate the expected value of the cloud distribution from the sample mean in Step 21:

[0139] Step 23: Calculate the entropy of the cloud distribution using the sample mean from Step 21 and the expected value of the cloud distribution from Step 22.

[0140] Step 24: The hyperentropy of the cloud distribution can be calculated from the sample variance in Step 21 and the entropy of the cloud distribution in Step 23.

[0141] Step 25: Calculate the cloud model digital features of the historical dataset and the test dataset using the methods from Steps 21 to 24. Then, calculate the similarity between the cloud model digital features of the historical dataset and the cloud model digital features of the test dataset, using the following formula:

[0142]

[0143] Where α=(Ex α En α He α ) represents the digital features of the cloud model in the historical dataset, β = (Ex β En β He β ) represents the digital features of the cloud model in the test dataset.

[0144] Furthermore, the local model construction module is specifically used to perform the following steps:

[0145] Step 31: Using the reservation data of train tickets for the designated scenic spot and the real-time search volume of the designated scenic spot on the electronic map as the sample input quantity p, and the number of visitors to the scenic spot as the sample output quantity q, there are p input quantities {x1, x2, ..., x}. p} and q output quantities {y1, y2, ..., y q Let} be the sample, and let the input data matrix X∈R be the sample. n×p Output data matrix Y∈R n×q Let the first principal component axis vectors of X and Y be w1(p×1) and c1(q×1), respectively, and both be unit vectors. Then w1 and c1 represent the first principal components t1 and u1 of X and Y, where t1 = X*w1 and u1 = Y*c1.

[0146] Step 32: Maximize the correlation between t1 and u1 using partial least squares method, as shown in the following formula:

[0147] Maximize<Xw1,Yc1> ,Subject to:||w1||=1,||c1||=1

[0148] Among them, Maximize<Xw1,Yc1> This represents maximizing the correlation between t1 and u1. t1 is the first component of X and the first axis of X. It is a unit vector, i.e., ||w1|| = 1. Similarly, we get ||c1|| = 1.

[0149] Step 33: Use the Lagrange multiplier method to find the weight vector w of the independent variable. k and the weight vector c of the dependent variable k , where w k is u'k v k v' k u k The eigenvector corresponding to the largest eigenvalue, c k v' k u k u' k v k The eigenvector corresponding to the largest eigenvalue; its independent variable is the latent variable t. k and the latent variable s of the dependent variable k The calculation is as follows:

[0150] t k =u k-1 w k

[0151] s k =v k-1 c k

[0152] Step 34: Extract the factor loadings p of the latent independent and dependent variables. k and q k and the regression coefficient b between the two k :

[0153] p' k =(t' k t k ) -1 t' k u k-1

[0154] q' k =(t' k t k ) -1 t' k v k-1

[0155] b' k =(t' k t k ) -1 t' k s k

[0156] Step 35: Calculate the residual space u of the independent variable x and the dependent variable y. k v k :

[0157] u k =u k-1 -t k p' k

[0158] v k =vk-1 -t k q' k

[0159] Step 36: Determine the residual space u k and v k Has the minimum value been reached? If not, return to step 32. If the minimum value has been reached, establish u. k and v k The linear regression equation between them:

[0160] v=uβ

[0161]

[0162] Step 36: Through inverse standardization, obtain the regression equations for the dependent variable Y and the independent variable X:

[0163] Y = Xβ.

[0164] Since the apparatus described in Embodiment 2 of the present invention is an apparatus used to implement the method of Embodiment 1 of the present invention, those skilled in the art can understand the specific structure and variations of the apparatus based on the method described in Embodiment 1 of the present invention, and therefore will not be described again here. All apparatuses used in the method of Embodiment 1 of the present invention fall within the scope of protection of the present invention.

[0165] While specific embodiments of the present invention have been described above, those skilled in the art should understand that the specific embodiments described are merely illustrative and not intended to limit the scope of the present invention. Equivalent modifications and variations made by those skilled in the art in accordance with the spirit of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for predicting the number of visitors in a scenic spot based on a big data platform, characterized in that, The method comprises the following steps: Step 10, based on a big data platform, real-time acquisition of reservation data of train tickets of a specified scenic spot and real-time search times of the specified scenic spot in an electronic map, integration of the reservation data of the train tickets and the real-time search times according to a time dimension to obtain a historical data set; Step 20, acquisition of current reservation data of train tickets of the specified scenic spot and real-time search times of the specified scenic spot in the electronic map as a test data set, similarity measurement of all samples in the historical data set and the test data set by a cloud model, and extraction of the first k samples as near-neighbor data of the test data set; Step 30, taking the number of tourists of the specified scenic spot at the next moment as a dependent variable, taking the near-neighbor data as independent variables, introducing a Lagrange multiplier to obtain a weight vector of the independent variables and a weight vector of the dependent variables, then extracting factors of latent variables of the independent variables and factors of latent variables of the dependent variables and a regression coefficient between the two, then calculating a residual space of the independent variables and the dependent variables, training a partial least squares model to minimize the residual space, and finally establishing a linear regression equation between the residual spaces and inversely normalizing to obtain a local model corresponding to the test data set; Step 40, input of the current reservation data of train tickets of the specified scenic spot and the real-time search times of the specified scenic spot in the electronic map into the local model to obtain an estimated value of the number of tourists of the specified scenic spot at the next moment.

2. The method of claim 1, wherein: In the step 20, the similarity measurement of all samples in the historical data set and the test data set by the cloud model specifically comprises: Step 21: calculating the reservation data of the train ticket of the specified scenic spot and the real-time search frequency quantitative value of the specified scenic spot in the electronic map sample mean first-order sample absolute central moment sample variance ; Step 22: Calculate the expected value of the cloud distribution from the sample mean in step 21: ; Step 23: Calculate the entropy of the cloud distribution from the sample mean in step 21 and the expected value of the cloud distribution in step 22: ; Step 24: The super entropy of the cloud distribution is calculated from the sample variance in step 21 and the entropy of the cloud distribution in step 23: Step 25, calculation of cloud model digital features of the historical data set and cloud model digital features of the test data set by the method of steps 21 to 24, and then calculation of the similarity of the cloud model digital features of the historical data set and the cloud model digital features of the test data, with the formula as follows: wherein, cloud model numeric features representing the historical data set, cloud model numeric features representing the test data set.

3. The method of claim 1, wherein: The step 30 specifically comprises: Step 31: Using the reservation data of train tickets for the designated scenic spot and the real-time search volume of the designated scenic spot on the electronic map as the sample input p, and the number of visitors to the scenic spot as the sample output q, there are p input quantities. and q output quantities For the sample, the input data matrix Output data matrix ;set up and The first principal component axis vectors are respectively and And since they are all unit vectors, then by and Indicate and The first principal component and ,in , ; Step 32, the correlation between the variables is maximized using a partial least squares method, as follows: and the correlation between the variables is maximized using a partial least squares method, as follows: wherein represents and maximizes the correlation between is the first component of X, is the first axis of X, which is a unit vector, i.e. has || = 1, and by analogy || = 1; Step 33, the method of introducing Lagrange multiplier to find the weight vector of independent variable and the weight vector of dependent variable where is the eigenvector corresponding to the largest eigenvalue of is the eigenvector corresponding to the largest eigenvalue of ; the latent variable of independent variable and the latent variable of dependent variable is calculated as Step 34, extracting factor loadings for independent latent variables and dependent latent variables and and regression coefficients between them : Step 35, computing the residual space of the independent variable x and the dependent variable y , : Step 36, determine residual space and whether a minimum is reached, if not return to step 32, if a minimum is reached, establish and a linear regression equation between Step 37, obtaining of a regression equation of the dependent variable Y and the independent variable X by inverse normalization: 。 4. A scenic spot reception number pre-judgment device based on a big data platform, characterized in that, The method comprises the following steps: A data preprocessing module, a cloud model similarity measurement module, a local model construction module, and a prediction output module; The data preprocessing module is configured to acquire, based on a big data platform, reservation data of train tickets of a specified scenic spot and real-time search times of the specified scenic spot in an electronic map, integrate the reservation data of the train tickets and the real-time search times according to a time dimension to obtain a historical data set; The cloud model similarity measurement module is configured to acquire current reservation data of train tickets of the specified scenic spot and real-time search times of the specified scenic spot in the electronic map as a test data set, measure the similarity of all samples in the historical data set and the test data set by a cloud model, and extract the first k samples as near-neighbor data of the test data set; The local model construction module is configured to take the number of tourists in the next moment of the specified scenic spot as the dependent variable, take the near-neighbor data as the independent variable, introduce a Lagrange multiplier to obtain a weight vector of the independent variable and a weight vector of the dependent variable, then extract factors of the independent variable latent variable and the dependent variable latent variable and a regression coefficient between the two, subsequently calculate a residual space of the independent variable and the dependent variable, train a partial least squares model to minimize the residual space, and finally establish a linear regression equation between the residual spaces and inverse-normalize the linear regression equation to obtain a local model corresponding to the test data set; The prediction output module is configured to input the current reservation data of the train ticket of the specified scenic spot and the real-time search times of the specified scenic spot in the electronic map into the local model to obtain an estimated value of the number of tourists in the next moment of the specified scenic spot.

5. The apparatus of claim 4, wherein: In the cloud model similarity measurement module, similarity between all samples in the historical data set and the test data set is measured by using the cloud model, and the method specifically includes the following steps: Step 21: calculating the reservation data of the train ticket of the specified scenic spot and the real-time search frequency quantitative value of the specified scenic spot in the electronic map sample mean first-order sample absolute central moment sample variance ; Step 22: Calculate the expected value of the cloud distribution from the sample mean in step 21: ; Step 23: Calculate the entropy of the cloud distribution from the sample mean in step 21 and the expected value of the cloud distribution in step 22: ; Step 24: The super entropy of the cloud distribution is calculated from the sample variance in step 21 and the entropy of the cloud distribution in step 23: Step 25: The cloud model digital features of the historical data set and the test data set are calculated according to the method of steps 21 to 24, and then the similarity between the cloud model digital features of the historical data set and the cloud model digital features of the test data is calculated, and the formula is as follows: wherein, cloud model numeric features representing the historical data set, cloud model numeric features representing the test data set.

6. The apparatus of claim 4, wherein: The local model construction module is specifically configured to perform the following steps: Step 31: Using the reservation data of train tickets for the designated scenic spot and the real-time search volume of the designated scenic spot on the electronic map as the sample input p, and the number of visitors to the scenic spot as the sample output q, there are p input quantities. and q output quantities For the sample, the input data matrix Output data matrix ;set up and The first principal component axis vectors are respectively and And since they are all unit vectors, then by and Indicate and The first principal component and ,in , ; Step 32, the correlation between the variables is maximized using a partial least squares method, as follows: and the correlation between the variables is maximized using a partial least squares method, as follows: wherein represents and maximizes the correlation between is the first component of X, is the first axis of X, which is a unit vector, i.e. has || = 1, and by analogy || = 1; Step 33, the method of introducing Lagrange multipliers to find the weight vector of the independent variable and the weight vector of the dependent variable where is the eigenvector corresponding to the largest eigenvalue of is the eigenvector corresponding to the largest eigenvalue of ; the latent variable of the independent variable and the latent variable of the dependent variable is calculated as: Step 34, extracting factor loadings for independent latent variables and dependent latent variables and and regression coefficients between them : Step 35, computing the residual space of the independent variable x and the dependent variable y , : Step 36, determine residual space and whether a minimum is reached, if not return to step 32, if a minimum is reached, establish and a linear regression equation between Step 37: The regression equation of the dependent variable Y and the independent variable X is obtained by inverse normalization. 。

Images