Connection mode selection prediction method based on MNL and BP neural network

Abstract

The invention discloses a connection mode selection prediction method based on an MNL and a BP neural network, and the method comprises the steps: obtaining the POI data of a region selected according to a determined to-be-predicted connection mode, determining a travel demand point, obtaining the real travel data of the region, building a data set, and carrying out the prediction of the connection mode selection based on a random forest model. According to the method, main influence factors of the connection comfort are screened and analyzed, connection mode selection is modeled based on MNL, MNL model parameters are identified by adopting a maximum likelihood estimation method, the connection mode selection probability is determined, an error correction model is constructed based on a BP neural network, and the connection comfort is improved. Training the error correction model based on the data determined by the MNL model and the data set; and finally, based on an error correction model of a BP neural network and an MNL model, obtaining a connection mode selection prediction result of the region. According to the prediction method, the error correction model based on the BP neural network is constructed, and the MNL model calculation value and the error correction value of the BP neural network error correction model are superposed to obtain the connection mode selection prediction model, so that the robustness of the connection mode selection probability prediction data can be ensured, and the prediction accuracy of the connection mode selection probability prediction model is improved. And prediction data can be ensured to have higher accuracy, and conditions are provided for sharing bicycle connection demand prediction.

Description

Technical Field

[0001] The invention belongs to the field of traffic planning and management, and in particular relates to a connection mode selection prediction method based on MNL and BP neural network. Background Art

[0002] In recent years, with the acceleration of urbanization and motorization, traffic congestion in large and medium-sized cities has become increasingly prominent. Due to the characteristics of "large capacity", "high punctuality" and "high reliability" of urban rail transit, urban rail transit has become an effective means to alleviate traffic congestion in major cities around the world. However, it is precisely the attributes of "large capacity" and "fixed routes" that determine that it cannot provide passengers with "door-to-door" transportation services. Therefore, the reasonable promotion of the connection between urban rail transit and other modes of transportation is an important issue that must be paid attention to in order to promote the sound development of urban transportation in my country.

[0003] In 2016, shared bicycles began to develop on a large scale in domestic cities. Their characteristics of "ride and use" and "wide distribution" greatly met the needs of residents for short-distance travel, and naturally became the most flexible connection option for urban rail transit. It not only provides a good solution to the "last mile" problem of urban rail transit, but also broadens the influence range of rail transit stations to a certain extent. Opportunities and challenges coexist. The imbalance between supply and demand and the accumulation of parking spots are particularly serious around subway stations during peak hours in the morning and evening. Therefore, how to accurately obtain the demand for cycling connection at rail stations, and then properly control the number of shared bicycles, and further improve their connection service level, has become a major challenge faced by governments and operating companies. At present, the methods for predicting the demand for cycling connection at rail transit stations are mainly divided into two categories. One is to first explore the factors affecting cycling connection through machine learning methods, and then predict the demand for cycling connection through neural networks or regression models. The other is to establish relevant mathematical models (such as multinomial logit models, MNL models) based on the influencing factors of the connection method selection and identify parameters to obtain the selection probability of different connection methods, thereby predicting the demand for different connection methods. Directly predicting the selection probability of different docking methods using only neural networks or regression models is prone to overfitting, which may lead to the predicted data being far from the actual situation, causing serious supply and demand imbalance, and the predicted data lacks robustness; although establishing relevant mathematical models based only on the influencing factors of docking method selection can greatly improve the robustness of the predicted data, the complexity of the actual situation and the comprehensiveness of the mathematical model will definitely be subject to certain limitations, which will lead to the predicted data not having a huge deviation from the actual data, but there will still be a certain deviation. In addition, the influencing factors considered in previous studies often ignore the comfort level, but the present invention has learned through preliminary questionnaires that comfort level is also a major factor affecting the choice of docking method. Therefore, it is necessary to implement a docking method selection prediction method that takes into account the comfort factor and takes into account both robustness and accuracy. Summary of the invention

[0004] In view of the above problems, the purpose of the present invention is to provide a connection mode selection prediction method based on MNL and BP neural network. When establishing the connection mode selection model for rail transit stations, in addition to considering the three major factors of connection distance, connection time and economic cost, the key factor of comfort is also considered. The probability calculation value obtained by using the MNL model is error corrected through the BP network, thereby achieving a robust and accurate estimation of the connection mode selection probability.

[0005] The specific technical solution for achieving the purpose of the present invention is:

[0006] A connection mode selection prediction method based on MNL and BP neural network includes the following steps:

[0007] Step 1: According to the area selected by the connection mode to be predicted, obtain the POI data of the area, determine the travel demand point, obtain the real travel data of the area, and build a data set;

[0008] Step 2: Based on the random forest model, the main influencing factors of docking comfort are screened and analyzed;

[0009] Step 3: Model the docking mode selection based on MNL, and use the maximum likelihood estimation method to identify the MNL model parameters to determine the probability of docking mode selection;

[0010] Step 4: construct an error correction model based on the BP neural network, and train the error correction model based on the data determined by the MNL model and the data set;

[0011] Step 5: Based on the BP neural network error correction model and the MNL model, the connection mode selection prediction results of the area are obtained.

[0012] Compared with the prior art, the present invention has the following beneficial effects:

[0013] (1) The prediction method of the present invention, in addition to considering the three major factors of connection distance, connection time, and economic cost, also considers the key factor of comfort when establishing a rail transit station connection mode selection model. Through the random forest model, the main influencing factors of connection comfort can be comprehensively considered;

[0014] (2) The prediction method of the present invention constructs an error correction model based on the BP neural network, and superimposes the calculated value of the MNL model with the error correction value of the BP network error correction model to obtain a prediction model for the docking mode selection. This can not only ensure the robustness of the probability prediction data for the docking mode selection, but also ensure that the prediction data has a higher accuracy, thus providing conditions for the prediction of shared bicycle docking demand.

[0015] The present invention is further described below in conjunction with specific implementation modes. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 The figure is a flow chart of the connection mode selection prediction method based on MNL and BP neural network of the present invention.

[0017] Figure 2 A schematic diagram of demand points in an area to be selected for a connection mode to be predicted in an embodiment of the present invention.

[0018] Figure 3This is a schematic diagram of connection mode selection prediction based on MNL and BP neural network in an embodiment of the present invention. DETAILED DESCRIPTION

[0019] Example

[0020] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention. The described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0021] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an kind" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0022] Unless otherwise specifically stated, the relative arrangement, numerical expressions and numerical values ​​of the parts and steps set forth in these embodiments do not limit the scope of the present application. At the same time, it should be understood that, for ease of description, the sizes of the various parts shown in the accompanying drawings are not drawn according to the actual proportional relationship. The technology, method and equipment known to those of ordinary skill in the relevant field may not be discussed in detail, but in appropriate cases, the technology, method and equipment should be considered as a part of the authorization specification. In all examples shown and discussed here, any specific value should be interpreted as being merely exemplary, rather than as a limitation. Therefore, other examples of exemplary embodiments may have different values. It should be noted that similar reference numerals and letters represent similar items in the following drawings, so that once a certain item is defined in an accompanying drawing, it does not need to be further discussed in subsequent drawings.

[0023] In response to the problem of imbalanced supply and demand of shared bicycles and accumulation of parking spots around subway stations during the morning rush hour, which has led to a decline in the level of docking service at rail transit stations, this application proposes a docking mode selection prediction method based on the MNL model and the BP network error correction model. This method can more comprehensively consider the influencing factors of different docking mode selections, and for the theoretical values ​​calculated using the MNL model, the error is corrected through BP network training, so that the obtained docking mode selection probability is more accurate, which lays a good foundation for improving the docking service level.

[0024] Combination Figure 1 , a connection mode selection prediction method based on MNL and BP neural network, comprising the following steps:

[0025] Step 1: According to the area selected by the connection mode to be predicted, obtain the POI data of the area, determine the travel demand point, obtain the real travel data of the area, and build a data set;

[0026] Specifically, the research area of ​​the general connection method is the rail transit station area, and the rail transit station I is selected to determine the research area range, and the research time period is determined to be the morning and evening peak hours of the day (7:00-9:00, 17:00-19:00). In this embodiment, the Xinglong Street Station of Nanjing Rail Transit Line 2 is selected as the research station I. Figure 2 As shown;

[0027] In terms of data acquisition, the acquisition method of this embodiment is: in the study area, an anonymous questionnaire survey is conducted on passengers waiting for the bus at the station for 30 consecutive working days to investigate RP (Revealed Preference) behavior. The survey content mainly includes passenger personal attributes, departure point, station entrance, connection selection method, departure time, station entry time, difficulty of parking when arriving at the station, whether the bus is on time, etc. Among them, in order to obtain the most accurate connection data as possible, the connection distance and connection time can be combined with the questionnaire information and given by calling the path planning algorithm with the help of the AutoNavi API open platform;

[0028] The collected questionnaires are screened for valid ones through indicators such as IP positioning and completion degree, and the POI data around the site is obtained to divide the study area into traffic districts, and the centroid of each district is selected as the travel demand point r, r∈{1,2,…,R}, such as Figure 2 As shown, various types of real travel data are obtained at the same time, where the distance between the travel demand point r and the subway station is the connecting distance.

[0029] Step 2: Based on the random forest model, the main influencing factors of docking comfort are screened and analyzed. Generally speaking, among the common docking methods, the comfort of walking docking is mainly related to the walking distance and the severity of the weather. The comfort of shared bicycle docking is mainly related to the difficulty of parking and the severity of the weather. The comfort of bus docking is mainly related to the distance between the departure point and the bus station and the waiting time.

[0030] Step 3: Model the docking mode selection based on MNL, and use the maximum likelihood estimation method to identify the MNL model parameters to determine the probability of docking mode selection:

[0031] Based on the MNL model, the results of step 2 screening are used as influencing factors to determine the prediction results of the connection mode selection based on the MNL model:

[0032]

[0033] Among them, r is the serial number of the travel demand point, r∈{1,2,…,R}, V j is the determinant item in the utility function of the passenger choosing the connection mode j, and the subscript m represents the corresponding influencing factor. is the parameter to be calibrated corresponding to the influencing factor m, is the value of influencing factor m when the passenger chooses connection mode j, P j is the probability of selecting connection mode j.

[0034] In this embodiment, the influencing factors include the connection distance X d , Connection time X t , connection fee X c , connection comfort X s , set "walking", "shared bicycle" and "bus" as connection modes 1, 2 and 3 respectively. For passengers in the same area, the utility function of connection mode j (j = 1, 2, 3) is:

[0035] U j =V j +ε j

[0036] Where V j is the deterministic term in the utility function of the passenger choosing connection mode j, ε j is the probability term in the utility function of the passenger choosing connection mode j.

[0037] V j It is a function composed of four factors that affect the passenger's choice of connection mode j to arrive at station I, as shown in the formula:

[0038]

[0039] Among them, V r j is the determinant item in the utility function of the passenger at the community demand point r choosing the connection mode j. The subscript m is the abbreviation of the corresponding four influencing factors. is the parameter to be calibrated corresponding to the influencing factor m, The value of influencing factor m corresponding to the passenger's choice of connection mode j;

[0040] The probability that passengers in this area choose the connection method is:

[0041]

[0042] In order to avoid exponential growth and increase the error, the formula is averaged, that is, each exponential term is divided by the average utility of each travel mode. The improved model is

[0043]

[0044] Then, the maximum likelihood estimation method is used to calibrate the parameters corresponding to the influencing factor m. Identify, that is, determine the value of each influencing factor m corresponding to different connection modes j based on the real data obtained And the connection ratio corresponding to different connection modes j is used as the predicted probability P j On this basis, the maximum likelihood estimation method is used to estimate the parameters corresponding to the influencing factor m under different connection methods j. To identify.

[0045] In addition, based on the real travel data obtained in this embodiment, the values ​​of each influencing factor are as follows:

[0046] (1) Connection distance X d , Connection time X t

[0047] Usually, the data obtained from questionnaire surveys are used. However, for passengers with uncertain travel time, the route planning algorithm is called based on the departure point and subway entrance information with the help of the Amap API open platform to give X d and X t , the units are km and min respectively.

[0048] (2) Connection fee X c

[0049] Define the connection fee X c It is the actual cost for passengers to travel from the departure point to station I. Walking transfer fee At present, there are two charging modes for shared bicycle rental: the first is 1.5 yuan for the first 15 minutes of a single ride, and the second is to buy a monthly card for about 20 yuan per month. Since the radius of the research area is only 1km, it can be assumed that the single ride time does not exceed 15 minutes. To simplify the calculation, it is assumed that the ratio of the number of people in these two rental modes is 1:1. The second mode is based on passengers riding twice on weekdays every month, and the cost of each ride is about 0.45 yuan, so the shared bicycle connection fee can be taken as The bus fare in Nanjing is a single ticket system, with a one-way fare of 2 yuan. Assuming that there is no bus transfer during the transfer process, the bus transfer fee is calculated. Yuan.

[0050] (3) Connection comfort X s

[0051] Through the survey on satisfaction with the docking mode and the reasons for it in the questionnaire, it was found that the comfort of walking is mainly related to the walking distance and the severity of the weather, the comfort of riding a shared bicycle is mainly related to the difficulty of parking and the severity of the weather, and the comfort of bus docking is mainly related to the distance between the starting point and the bus station, the waiting time, and the severity of the weather, and all of the above are negatively correlated. Assuming that the full score of comfort is 2, and the penalty factor β = 0.1 is introduced, the docking comfort score of each mode is as follows

[0052]

[0053]

[0054]

[0055] In the formula, d1 = 0.787 is the comfortable walking distance, d3 = 0.252 is the comfortable distance between the starting point and the bus stop, and X d3 is the distance between the starting point and the bus stop, w 1,2,3 They represent the severity of the weather (good, average, bad), the difficulty of parking (easy, not easy, relatively difficult), and the waiting time for the bus (short, just acceptable, long). They are divided into three levels and assigned scores of 0, 1, and 2 according to the degree to determine whether to add a penalty factor.

[0056] Step 4: construct an error correction model based on the BP neural network, and train the error correction model based on the data determined by the MNL model and the data set, such as Figure 3 As shown;

[0057] During the training process, the influencing factors screened out in step 2 are used as network input, and the difference between the probability of selecting each connection mode determined based on the MNL model and the actual probability of selecting each connection mode determined based on real travel data is used as network output. The BP neural network is trained so that the BP neural network can correct the predicted probability error value.

[0058] Step 5: Based on the BP neural network error correction model and the MNL model, obtain the prediction results of the connection mode selection in the area:

[0059]

[0060] in, Indicates the connection mode selection prediction result value of the area to be predicted. It represents the error correction value of the connection mode selection prediction probability of the area to be predicted output by the error correction model.

[0061] The present invention also provides a connection mode selection prediction system based on MNL and BP neural network, comprising the following steps:

[0062] Data acquisition module: used to obtain POI data of the area selected according to the determined connection mode to be predicted, determine the travel demand point, and obtain the real travel data of the area to construct a data set;

[0063] Influencing factor determination module: used to screen and analyze the main influencing factors of docking comfort based on the random forest model;

[0064] Connection mode selection probability calculation module: used to model the connection mode selection based on MNL, identify the MNL model parameters using the maximum likelihood estimation method, and determine the connection mode selection probability;

[0065] Error correction module: used to build an error correction model based on BP neural network, and train the error correction model based on the data determined by the MNL model and the data set;

[0066] Connection mode selection prediction module: used for the error correction model based on BP neural network and MNL model to obtain the connection mode selection prediction results of the area.

[0067] A computer device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the following steps are implemented:

[0068] Step 1: According to the area selected by the connection mode to be predicted, obtain the POI data of the area, determine the travel demand point, obtain the real travel data of the area, and build a data set;

[0069] Step 2: Based on the random forest model, the main influencing factors of docking comfort are screened and analyzed;

[0070] Step 3: Model the docking mode selection based on MNL, and use the maximum likelihood estimation method to identify the MNL model parameters to determine the probability of docking mode selection;

[0071] Step 4: construct an error correction model based on the BP neural network, and train the error correction model based on the data determined by the MNL model and the data set;

[0072] Step 5: Based on the BP neural network error correction model and the MNL model, the connection mode selection prediction results of the area are obtained.

[0073] A computer storable medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the following steps are implemented:

[0074] Step 1: According to the area selected by the connection mode to be predicted, obtain the POI data of the area, determine the travel demand point, obtain the real travel data of the area, and build a data set;

[0075] Step 2: Based on the random forest model, the main influencing factors of docking comfort are screened and analyzed;

[0076] Step 3: Model the docking mode selection based on MNL, and use the maximum likelihood estimation method to identify the MNL model parameters to determine the probability of docking mode selection;

[0077] Step 4: construct an error correction model based on the BP neural network, and train the error correction model based on the data determined by the MNL model and the data set;

[0078] Step 5: Based on the BP neural network error correction model and the MNL model, the connection mode selection prediction results of the area are obtained.

[0079] In this embodiment, the connection data of Xinglong Street Station from the 1st to the 29th day of the survey is used to establish a prediction model, and the connection data on the 30th day is used to verify the above model. The model verification results are shown in Table 1 and Table 2. The results show that the model has good accuracy, and the corrected probability value is closer to the actual value and has higher accuracy. Therefore, this method can be used to predict the probability of connection mode selection and provide effective data for the dynamic adjustment of urban transportation planning, which is conducive to the implementation of urban intelligent transportation and the improvement of citizens' happiness.

[0080] Table 1. Validation results of morning peak model

[0081]

[0082] Table 2. Validation results of evening peak model

[0083]

[0084] When establishing a rail transit station docking mode selection model, the present invention not only considers the three major factors of docking distance, docking time, and economic cost, but also considers the key factor of comfort. The probability calculation value obtained using the MNL model is error corrected through the BP network, thereby achieving a robust and accurate estimation of the docking mode selection probability.

[0085] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A connection mode selection prediction method based on MNL and BP neural network, characterized in that: The following steps are involved: Step 1: According to the area selected by the connection mode to be predicted, obtain the POI data of the area, determine the travel demand point, obtain the real travel data of the area, and build a data set; Step 2: Based on the random forest model, the main influencing factors of docking comfort are screened and analyzed; Step 3: Model the docking mode selection based on MNL, and use the maximum likelihood estimation method to identify the MNL model parameters to determine the probability of docking mode selection; Step 4: construct an error correction model based on the BP neural network, and train the error correction model based on the data determined by the MNL model and the data set; Step 5: Based on the BP neural network error correction model and the MNL model, the connection mode selection prediction results of the area are obtained.

2. The connection mode selection prediction method based on MNL and BP neural network according to claim 1 is characterized in that: The selection of the docking method is modeled based on the MNL model in step 3, specifically: Based on the MNL model, the results of step 2 screening are used as influencing factors to determine the prediction results of the connection mode selection based on the MNL model: Among them, r is the serial number of the travel demand point, r∈{1,2,…,R}, V j is the determinant item in the utility function of the passenger choosing the connection mode j, and the subscript m represents the corresponding influencing factor. is the parameter to be calibrated corresponding to the influencing factor m, is the value of influencing factor m when the passenger chooses connection mode j, P j is the probability of selecting connection mode j.

3. The connection mode selection prediction method based on MNL and BP neural network according to claim 2 is characterized in that: The influencing factors include the connection distance X d , Connection time X t , connection fee X c , connection comfort X s .

4. The connection mode selection prediction method based on MNL and BP neural network according to claim 2 is characterized in that: The maximum likelihood estimation method is used in step 3 to identify the parameters of the MNL model, specifically: Determine the value of each influencing factor m corresponding to different connection modes j based on the real data obtained And the connection ratio corresponding to different connection modes j is used as the predicted probability P j On this basis, the maximum likelihood estimation method is used to estimate the parameters corresponding to the influencing factor m under different connection methods j. To identify.

5. The connection mode selection prediction method based on MNL and BP neural network according to claim 2 is characterized in that: During the training process in step 4, the influencing factors screened out in step 2 are used as network input, and the difference between the probability of selecting each connection mode determined based on the MNL model and the actual probability of selecting each connection mode determined based on real travel data is used as output to train the BP neural network and output the probability error correction value predicted by the BP neural network.

6. The connection mode selection prediction method based on MNL and BP neural network according to claim 2 is characterized in that: The connection mode selection prediction result in step 5 is specifically: in, Indicates the connection mode selection prediction result value of the area to be predicted. It represents the error correction value of the connection mode selection prediction probability of the area to be predicted output by the error correction model.

7. A connection mode selection prediction system based on MNL and BP neural network, characterized in that: The following steps are involved: Data acquisition module: used to obtain POI data of the area selected according to the determined connection mode to be predicted, determine the travel demand point, and obtain the real travel data of the area to construct a data set; Influencing factor determination module: used to screen and analyze the main influencing factors of docking comfort based on the random forest model; Connection mode selection probability calculation module: used to model the connection mode selection based on MNL, identify the MNL model parameters using the maximum likelihood estimation method, and determine the connection mode selection probability; Error correction module: used to build an error correction model based on BP neural network, and train the error correction model based on the data determined by the MNL model and the data set; Connection mode selection prediction module: used for the error correction model based on BP neural network and MNL model to obtain the connection mode selection prediction results of the area.

8. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer storable medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images