CN114328675A - Prediction method of bus travel time based on dual attention mechanism and bidirectional double layer LSTM - Google Patents

Abstract

The invention relates to a bus travel time prediction method based on a double-attention mechanism and a bidirectional double-layer LSTM, comprising the following steps of S1, acquiring bus attribute data and station entering and exiting data, and constructing a travel time basic data set; s2, acquiring basic characteristics, running time and weather characteristics of the bus during running, and constructing a travel time characteristic data set; step S3, based on correlation coefficient and variance analysis, carrying out correlation analysis on the characteristic factors in the travel time characteristic data set and the travel time, eliminating the characteristic factors which are irrelevant and have poor correlation, and matching with the travel time basic data set to obtain a travel time prediction data set; and S4, constructing a double-attention machine mechanism and a bidirectional double-layer LSTM neural network bus travel time prediction model, inputting the travel time prediction data set into the model, predicting the travel time of the bus, and outputting the predicted travel time of the bus. The invention can accurately predict the travel time of the bus and realize high-efficiency bus scheduling.

Description

Prediction method of bus travel time based on dual attention mechanism and bidirectional double layer LSTM

technical field

The invention relates to the field of urban intelligent transportation, in particular to a bus travel time prediction method based on a double attention mechanism and a two-way double-layer LSTM.

Background technique

As an important part of public transportation, buses have the advantages of small area per capita, low energy consumption per capita, large carrying capacity, and low overall transportation cost. By improving the service level of buses and attracting more people to use public transportation, it will help alleviate urban traffic congestion, reduce urban air pollution, and ensure the healthy and sound operation of the city. Therefore, vigorously developing buses is an important measure to realize the sustainable development of urban transportation and promote the virtuous circle and interactive development between people and society.

At present, bus scheduling is mainly estimated by dispatchers from their own experience, or based on traditional bus travel time prediction models (such as historical data models, time series models, regression prediction models, support vector machine models, Kalman filter models, etc. ) to get the estimated travel time. The estimation error of this method is often large, and the scheduling is often not timely enough, resulting in the phenomenon of "cross train" or "large interval", which in turn leads to low operating efficiency of the bus, poor service level, low attractiveness, and low utilization rate. far from ideal.

SUMMARY OF THE INVENTION

In view of this, the purpose of the present invention is to provide a bus travel time prediction method based on dual attention mechanism and bidirectional double-layer LSTM, which can more accurately estimate the travel time of buses and realize efficient bus scheduling.

To achieve the above object, the present invention adopts the following technical solutions:

A bus travel time prediction method based on dual attention mechanism and bidirectional double layer LSTM, including the following steps:

Step S1: obtain bus attribute data and inbound and outbound data, and preprocess to construct a basic data set of travel time;

Step S2: obtain the basic characteristics of the bus, the running time and the weather characteristics during running, and construct the travel time characteristic data set;

Step S3: based on the correlation coefficient and variance analysis, carry out a correlation analysis on the characteristic factors in the travel time characteristic data set and the travel time, discard the characteristic factors that are irrelevant and poorly correlated, and match with the travel time basic data set to obtain the travel time. time prediction dataset;

Step S4: Build a bus travel time prediction model with a dual attention mechanism and a bidirectional double-layer LSTM neural network, input the travel time prediction data set into the model, predict the travel time of the bus, and output the estimated travel time of the bus.

Further, the step S1 is specifically:

Step S11: obtain bus attribute data and entry and exit data, directly remove duplicate records and abnormal data and complete missing data;

Step S12: Match the bus attribute data with the inbound and outbound data according to the bus operation shift number, obtain the bus running time of each operation shift, and construct a basic data set of travel time.

Further, the step S2 is specifically: acquiring the vehicle number, driver number, vehicle departure time, and departure interval of the bus; acquiring a similar time period of the previous day and a similar time period of the same characteristic day of the previous week from the starting station to the end point. The running time of the station is obtained; the weather characteristics of the vehicle during operation include weather type, humidity, wind speed, and temperature, and a data set of travel time characteristics is constructed.

Further, the step S3 is specifically:

Step S31: using the Pearson correlation coefficient correlation coefficient to analyze the correlation between the continuous characteristic factor and the bus running time, and obtain the correlation between the variables;

Step S32: classify and encode discrete characteristic factors such as weather type, driver number, bus number, etc., convert them into type-specific characteristic factors, and use variance test to analyze the correlation between the type-specific characteristic factors and the bus running time;

Step S33: Discard irrelevant and poorly correlated feature factors, match the reserved features with the travel time basic data set, and obtain the travel time prediction data set.

Further, the step S4 is specifically:

Step S41: input the travel time prediction data set into the feature importance extraction module, assign different degrees of attention to the feature according to the difference in importance of the feature, and obtain the feature importance matrix;

Step S42: splicing the feature importance matrix with the input features of the model to obtain a bus operating state matrix;

Step S43: Input the bus running state matrix into the time feature extraction module, and use the bidirectional LSTM to extract the time features during the bus running process to generate a running time feature matrix;

Step S44: Input the obtained running time feature matrix into the travel time prediction module, fuse the attention into the time step of LSTM, build the Attention_LSTM layer, and realize the feature extraction of the influence on buses of different time distances; in the fully connected layer , using the root mean square error as the loss function, comparing the predicted results with the actual results, and continuously training the model to predict the travel time of the bus.

A bus travel time prediction system based on dual attention mechanism and bidirectional double layer LSTM, including

The feature importance extraction module uses the attention mechanism to obtain the importance weight of each feature to generate a feature importance matrix, distinguish the influence of different features on the bus travel time, and improve the prediction efficiency and prediction accuracy of the model;

The temporal feature extraction module uses the strong temporal feature capture capability of the bidirectional LSTM to capture the running features of the buses running before and after, and obtain the temporal features during the running process of the buses;

The travel time prediction module introduces attention into LSTM, builds the Attention_LSTM layer, and gives different attention to different time steps in LSTM, showing the impact of vehicles at different time distances on bus travel time.

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

The present invention introduces LSTM into the bus travel time prediction. On the one hand, the bidirectional LSTM is used to capture the time characteristics of the front and rear buses during the running process; The impact feature extraction of the model effectively improves the prediction accuracy of the model; realizes more accurate estimated bus travel time and efficient bus scheduling.

Description of drawings

Fig. 1 is the schematic flow chart of the method of the present invention;

Figure 2. Structure diagram of bus travel time prediction model with dual attention mechanism and bidirectional double-layer LSTM;

Fig. 3 is a bus driving route diagram of B2 bus road in the embodiment of the present invention;

FIG. 4 is a comparison diagram of the travel time prediction result obtained in the embodiment of the present invention and the actual value.

Detailed ways

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

Referring to Fig. 1, the present invention provides a method for predicting bus travel time based on a dual attention mechanism and a bidirectional double-layer LSTM, comprising the following steps:

In step S1, the attribute data and entry and exit data of the bus are obtained, and duplicate records and abnormal data are directly eliminated, and the missing data is filled. According to the bus operation shift number, match the bus attribute data with the inbound and outbound data, obtain the bus running time of each operating shift, and build a basic data set of travel time;

Step S2: Obtain basic features such as vehicle number, driver number, vehicle departure time, departure interval, etc. of the bus; obtain the running time from the starting station to the destination station in a similar time period of the previous day and the similar time period of the same characteristic day of the previous week etc.; obtain weather characteristics such as weather type, humidity, wind speed, temperature, etc. when the vehicle is running, and construct a travel time characteristic data set;

Step S3, based on the correlation coefficient and variance analysis, perform a correlation analysis on the characteristic factors in the characteristic data set and the travel time, discard the characteristic factors that are irrelevant and poorly correlated, and match with the basic data set of travel time to obtain travel time predictions data set;

Step S4, build a dual-attention mechanism and a bidirectional double-layer LSTM neural network bus travel time prediction model based on the attention mechanism and LSTM, input the travel time prediction data set into the model, predict the travel time of the bus, and output the bus The estimated travel time of the car.

Example 1:

Referring to Figure 3, this embodiment takes the bus operating vehicles of B2 Road in Guangzhou as the research object, and selects the B2 road during the evening rush hour (17:00-19:00) on weekdays from October 1, 2020 to December 11, 2020 The bus data is used as instance data, and the specific implementation is as follows:

In step S1, the attribute data and entry and exit data of the bus are obtained, and duplicate records and abnormal data are directly eliminated, and the missing data is filled. According to the bus operation shift number, the bus attribute data is matched with the inbound and outbound data, and the bus running time of each operating shift is obtained, and the basic data set of travel time is constructed.

1. Collect data. The collected experimental data includes two parts: the first part is static data, including the basic information of the platform on the B2 bus line; the second part is dynamic data, including B2 bus entry and exit data, bus attribute data.

The basic information of the platform includes the platform number, the latitude and longitude of the platform, the up and down identification of the platform, etc.; the bus entry and exit data includes the road list ID, the vehicle number, the platform number, the platform name, the entry time, the exit time, the up and down identification, the date, etc.; The bus attribute data includes road list ID, departure time, vehicle number, driver number, route type, departure interval, etc.

2. Data preprocessing. Data preprocessing mainly includes bus entry and exit data preprocessing, bus attribute data preprocessing, static data and dynamic data matching and travel time acquisition.

For the preprocessing of bus scheduling data, the main purpose is to eliminate the repeated recorded data and extract the whole operation data.

Preprocessing for bus entry and exit data, including removing duplicate data, calibrating abnormal data, interpolating missing data, and obtaining bus travel time. First, the inbound and outbound data of the repeated records are directly eliminated; secondly, the abnormal data is processed. The abnormal data refers to the data with time-down and running time far less than the average station running time, which is first eliminated and then filled with interpolation; Third, the missing data is processed, mainly focusing on the missing data of the terminal arrival time. By calculating the proportion of the average travel time and average parking time of each part in the missing station, the interpolation time of each part is calculated in the missing time period, and the completion is performed. bus travel time.

Static data and dynamic data match. According to the bus operation shift number, the platform basic information, bus attribute data and bus entry and exit data are matched to obtain the basic data set of travel time.

Step S2: Obtain basic features such as vehicle number, driver number, vehicle departure time, departure interval, etc. of the bus; obtain the running time from the starting station to the destination station in a similar time period of the previous day and the similar time period of the same characteristic day of the previous week etc.; obtain weather characteristics such as weather type, humidity, wind speed, and temperature when the vehicle is running, and construct a travel time feature dataset.

Step S3, based on the correlation coefficient and variance analysis, perform a correlation analysis on the characteristic factors in the characteristic data set and the travel time, discard the characteristic factors that are irrelevant and poorly correlated, and match with the basic data set of travel time to obtain travel time predictions data set.

1. Use the Pearson coefficient correlation coefficient to analyze continuous features (vehicle departure time, departure interval, travel time in a similar time period on the previous day of the operating shift, travel time on the same characteristic day in the previous week, humidity, wind speed, temperature) and bus Correlation analysis of the travel time of the car.

2. Use variance analysis to perform variance analysis on classification features (bus vehicle number, driver number, weather type) and bus travel time.

3. Obtain the travel time prediction data set according to the correlation coefficient and variance analysis results.

Step S4, based on the attention mechanism and LSTM, construct a "dual attention mechanism and bidirectional double-layer LSTM neural network bus travel time prediction model", input the travel time prediction data set into the model, and predict the travel time of the bus, Output the estimated travel time for the bus.

1. Build a bus travel time prediction model based on the Kares deep learning framework with dual attention mechanism and bidirectional double-layer LSTM.

2. Divide the travel time prediction data set, and use the bus operation data of B2 during the evening rush hour on weekdays from October 1, 2020 to December 4, 2020 for training; use December 7 to December 11 The operation data of the B2 bus in the evening rush hour of the working day is used for verification.

3. Input the training data in the travel time prediction data set into the bus travel time prediction model of the dual attention mechanism and the bidirectional double-layer LSTM for training, using the root mean square error as the measurement index.

4. Predict the travel time of the B2 bus in the evening rush hour on weekdays from December 7th to December 11th, and output the estimated travel time.

Following the above specific implementation steps, a comparison chart of the predicted value and the actual value of the bus running time during the weekday evening peak from December 7 to December 11 is obtained. Based on the predicted time of the bus travel time predicted by the present invention, the average absolute error is 6.23%.

Table 1

As shown in Table 1, Table 1 compares the prediction results of the evening peak of each weekday of the dual-attention bi-directional double-layer LSTM model;

From Monday to Friday, there are differences in the expressiveness of the model. From Monday to Thursday, it has a high prediction accuracy, and its average error is lower than the total average error. The prediction accuracy is the best on Monday, and the average absolute error is about 3 minutes. However, the prediction accuracy on Friday dropped significantly, with an average absolute percentage error of only 8.88% and an average absolute error of around 7 minutes. Through the analysis of the B2 bus line, it can be found that the B2 bus line has two characteristics. Feature 1, there are 7 colleges and universities that pass through many B2 routes, namely: Guangzhou University of Technology, Guangzhou Institute of Engineering and Technology, Guangzhou Institute of Physical Education, South China Normal University, Jinan University, Guangdong Institute of Posts and Telecommunications and Guangdong Normal University of Technology; Feature 2 , There are 16 department stores and shopping centers with more routes on the B2 route, namely: Timeli, Laibai Plaza, Xindaxin Department Store, Guangbai Department Store, VT101 Victoria Plaza, Tee Mall, Grandview Plaza, Wanhua Linghui, Taikoo Hui, Modern Department Store, Tianyu Plaza, Wal-Mart, Juntang Shopping Plaza, Tee Mall, Constellation Plaza, Dongpu Shopping Center. Therefore, during the rush hour on Friday evening, more college students and citizens went to major department stores and shopping centers to play, and the number of trips increased significantly compared with Monday to Thursday. In this case, on the one hand, the number of people getting on and off the bus has increased significantly, the bus stop time at each station has changed, and the complexity of the time and space rules of operation has increased; As a result, the road conditions of the operating routes where the bus is located have become more complicated. Therefore, it is difficult for the model to predict the travel time of bus vehicles during the rush hour on Friday night, and it is difficult to capture the spatiotemporal regularity of bus operation in a complex traffic environment. The accuracy is reduced, but the average error is still low. At 9%, it generally meets the bus scheduling needs of bus companies.

In this embodiment, the dual-attention bidirectional double-layer LSTM model is compared with other neural network models experimentally.

Table 2

Specifically, in order to better analyze the improvement effect of the feature importance extraction module, the time feature module and the travel time prediction module on the travel time prediction accuracy, the prediction accuracy of the model is compared from two aspects. On the one hand, the model is compared with other neural network models: compared with the single-attention bidirectional double-layer LSTM model, the relative accuracy is improved by 7.98%; compared with the non-attention bidirectional double-layer LSTM model, the relative accuracy is improved by 16.3% ; Compared with the two-layer LSTM, the relative accuracy is improved by 19.3%. On the other hand, compared with the travel time prediction model of the traditional method: compared with the multilayer perceptron, the relative accuracy is improved by 29.0%; compared with the support vector machine model, the relative accuracy is improved by 31.0%. From the comparison results, it can be seen that the feature importance extraction module and the time feature module have a great improvement in improving the accuracy of bus travel time prediction. Incorporated into model predictions to obtain higher prediction accuracy.

The above descriptions are only preferred embodiments of the present invention, and all equivalent changes and modifications made according to the scope of the patent application of the present invention shall fall within the scope of the present invention.

Claims (6)

1. a method for predicting bus travel time based on double attention mechanism and two-way double-layer LSTM, is characterized in that, comprises the steps: Step S1: obtain bus attribute data and inbound and outbound data, and preprocess to construct a basic data set of travel time; Step S2: obtain the basic characteristics of the bus, the running time and the weather characteristics during running, and construct the travel time characteristic data set; Step S3: Based on the correlation coefficient and variance analysis, correlation analysis is performed on the characteristic factors in the travel time characteristic data set and the travel time, the characteristic factors that are irrelevant and poorly correlated are discarded, and matched with the travel time basic data set to obtain the travel time. time prediction dataset; Step S4: Build a bus travel time prediction model with a dual attention mechanism and a bidirectional double-layer LSTM neural network, input the travel time prediction data set into the model, predict the travel time of the bus, and output the estimated travel time of the bus. 2. the bus travel time prediction method based on dual attention mechanism and bidirectional double layer LSTM according to claim 1, is characterized in that, described step S1 is specifically: Step S11: obtain bus attribute data and entry and exit data, directly remove duplicate records and abnormal data and complete missing data; Step S12: Match the bus attribute data with the inbound and outbound data according to the bus operation shift number, obtain the bus running time of each operation shift, and construct a basic data set of travel time. 3. The bus travel time prediction method based on dual attention mechanism and two-way double-layer LSTM according to claim 1, is characterized in that, described step S2 is specifically: obtain the vehicle number, driver number, vehicle number of the bus Departure time and departure interval; obtain the running time from the starting station to the destination station in a similar time period of the previous day and the similar time period of the same characteristic day of the previous week; obtain the weather characteristics of the vehicle during operation, including weather type, humidity, wind speed, temperature, Build a travel time feature dataset. 4. the bus travel time prediction method based on double attention mechanism and bidirectional double layer LSTM according to claim 1, is characterized in that, described step S3 is specifically: Step S31: using the Pearson correlation coefficient correlation coefficient to analyze the correlation between the continuous characteristic factor and the bus running time, and obtain the correlation between the variables; Step S32: classify and encode the discrete characteristic factors of weather type, driver number, and bus number, convert them into type-specific characteristic factors, and use variance test to analyze the correlation between the type-specific characteristic factors and the bus running time; Step S33: Discard irrelevant and poorly correlated feature factors, match the reserved features with the travel time basic data set, and obtain the travel time prediction data set. 5. The bus travel time prediction method based on dual attention mechanism and two-way double-layer LSTM according to claim 1, is characterized in that, described step S4 is specifically: Step S41: input the travel time prediction data set into the feature importance extraction module, assign different degrees of attention to the feature according to the difference in importance of the feature, and obtain the feature importance matrix; Step S42: splicing the feature importance matrix with the input features of the model to obtain a bus operating state matrix; Step S43: Input the bus running state matrix into the time feature extraction module, and use the bidirectional LSTM to extract the time features during the bus running process to generate a running time feature matrix; Step S44: Input the obtained running time feature matrix into the travel time prediction module, fuse the attention into the time step of LSTM, build the Attention_LSTM layer, and realize the feature extraction of the influence on buses of different time distances; in the fully connected layer , using the root mean square error as the loss function, comparing the predicted results with the actual results, and continuously training the model to predict the travel time of the bus. 6. A bus travel time prediction system based on dual attention mechanism and bidirectional double-layer LSTM, characterized in that it includes The feature importance extraction module uses the attention mechanism to obtain the importance weight of each feature to generate a feature importance matrix, distinguish the influence of different features on the bus travel time, and improve the prediction efficiency and prediction accuracy of the model; The temporal feature extraction module uses the strong temporal feature capture capability of the bidirectional LSTM to capture the running features of the buses running before and after, and obtain the temporal features during the running process of the buses; The travel time prediction module introduces attention into the LSTM, builds the Attention_LSTM layer, and gives different attention to different time steps in the LSTM, showing the impact of vehicles at different time distances on the travel time of the bus.

Images