Method and device for predicting order completion amount of online car-hailing order based on CTCN (China Train Control Network)
By building a network model based on CTCN, combining CNN and TCN, using driver, environment and activity incentive characteristics, the accuracy and speed of online car-hailing order completion prediction are solved, and operational efficiency and driver activity are improved.
Patent Information
- Application Number
- CN202510466661.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-15
- Publication Date
- 2025-08-15
AI Technical Summary
It is difficult for the existing technology to accurately predict the number of online car-hailing orders, resulting in driver loss and reduced platform operation efficiency.
A network model based on CTCN is adopted, combined with CNN and TCN, a time series of drivers is constructed, and drivers, environment and activity motivation characteristics are used for prediction. Through data collection, feature integration and model training, the number of orders completed next day is predicted.
It improves the accuracy and speed of the order quantity prediction, conforms to the individual driver's situation, solves the problem of inaccurate regression prediction caused by sparse data, and enhances the effectiveness of the operation strategy.
Smart Images
Figure CN120495055A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of data processing technology, and more specifically, to a method and device for predicting the number of completed online ride-hailing orders based on CTCN. Background Art
[0002] With the advancement of mobile Internet technology and the popularization of smart phones, the online ride-hailing industry has rapidly emerged and gradually become an important part of urban transportation. The market competition is fierce, and various online ride-hailing platforms are vying to attract passengers and drivers, increase the number of completed orders and driver activity, in order to improve their market share and customer satisfaction, and optimize operational efficiency.
[0003] In the ride-hailing industry, marketing campaigns are a crucial tool for attracting and retaining drivers. Rewarding drivers for completing a specific number of orders effectively increases driver activity and willingness to accept orders, thereby improving overall driver completion volume and service quality. Driver completion volume directly reflects driver efficiency and income, influencing driver motivation. Low completion volume can lead to driver churn, impacting the platform's overall operational efficiency. Increasing completion volume not only impacts individual driver income but also the service quality and market reputation of the entire ride-hailing platform. Therefore, accurately forecasting completion volume is crucial for developing effective operational strategies. Summary of the Invention
[0004] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a method and device for predicting the number of completed online car-hailing orders based on CTCN to solve the problems raised in the above-mentioned background technology.
[0005] To achieve the above objectives, the present invention provides the following technical solution: a method for predicting the number of completed online ride-hailing orders based on CTCN, which specifically includes the following steps:
[0006] Step S1: extract the order data of the past 180 days from the database, perform data preprocessing and store it in the data warehouse;
[0007] Step S2: extracting driver characteristics, environmental characteristics, and activity incentive characteristics from the data warehouse, and integrating the extracted characteristics in the "driver-day" dimension to construct a driver time series;
[0008] Step S3: Segment the driver's time series, use the segmented time series as input, and the next-day completed orders in the series as output, build a CTCN network model, and train the CTCN network model;
[0009] Step S4: Use the trained CTCN network model to make predictions, store the daily feature data in the data warehouse, extract the feature data of the target driver, input it into the CTCN network model, and predict the number of orders completed the next day.
[0010] In a preferred embodiment, in step S1, order data for the past 180 days is extracted from the database, and data preprocessing is performed to exclude dirty data, test data, and erroneous data to ensure the authenticity and accuracy of the data, and the data is stored in the data warehouse. The specific steps are as follows:
[0011] Step A1, Data Extraction: Establish a connection with the source database and extract order data from the past 180 days, including order ID, driver ID, order date, and status information;
[0012] Step A2, Data Storage: Check the extracted order data, handle missing values and outliers, convert date fields to a standard date format, create a new table in the data warehouse to store the processed order data, and load the cleaned and processed data into the target table.
[0013] In a preferred embodiment, in step S2, driver characteristics, environmental characteristics, and activity incentive characteristics are extracted from the data warehouse, and the extracted characteristics are integrated in the "driver-day" dimension to construct a driver time series. The specific steps are as follows:
[0014] Step B1, Feature Extraction: Connect to the data warehouse to query and extract feature data related to drivers, environment, and activity incentives. The driver features include each driver's daily records, including whether the driver was effectively dispatched on that day, the duration of the dispatch, and the number of completed orders. Statistical data from the past 7 days and 30 days are summarized, including the number of effective dispatch days, total dispatch duration, and number of completed orders. The environmental features include the number of dispatched drivers in the city on that day, the average monthly number of dispatched drivers, the average number of completed orders, as well as the daily order-to-order ratio, order matching rate, response rate, and order completion rate. The activity incentive features include the driver's activity level and reward amount on that day. If there is no activity, the level is 0 and the reward is 0. Based on the number of completed orders on that day, the next day's activity level, reward amount, number of orders to be completed, and the increase in the level reward are predicted.
[0015] Step B2, Feature Integration: Create a data table structure with the "driver-day" dimension for each driver, ensuring that each driver has a corresponding record for each day. The extracted driver characteristics, environmental characteristics, and activity incentive characteristics are then integrated according to the "driver-day" dimension to ensure that each driver's characteristic data for each day is correctly associated.
[0016] Step B3, construct time series: using "driver-day" as the dimension, arrange the integrated feature data in chronological order and construct a time series data table X = (x1, x2, ..., x n ), where x n It is the feature vector of the nth day. Each row represents the complete feature record of a driver on a certain day. The constructed time series data is saved in the data warehouse for subsequent analysis and model training.
[0017] In a preferred embodiment, in step S3, the driver's time series is segmented, the segmented time series is used as input, and the next-day completed orders in the series are used as output to construct a CTCN network model and train the CTCN network model. The specific steps are as follows:
[0018] Step C1: Split the time series: Extract the time series feature data of drivers from the data warehouse, traverse the entire time series, and split it into multiple segments at 30-day intervals, with each segment containing 30 days of data. For each segment, use the feature data in the segment as input, and output the number of completed orders the next day after the last day of the segment;
[0019] Step C2: Constructing a CTCN network model: The CTCN network model is a new network architecture that combines the CNN network and the TCN network. CNN is a convolutional neural network used to extract features and model spatial relationships in data, and TCN is a temporal convolutional network used to capture dynamic changes in time series data. The model further includes the following steps:
[0020] Step C201: Define the model structure: The input layer accepts 30 days of time series data, uses a one-dimensional convolutional layer to extract spatial features, adds a temporal convolutional layer to capture the dynamic changes of the time series, uses a pooling layer to reduce the dimensionality of the data, uses a fully connected layer to process the features extracted by the convolutional layer, maps them to the final prediction result, and finally, the output layer outputs the prediction result of the number of orders completed for the next day;
[0021] Step C202, convolution operation: the feature vector of each driver at different time steps is expressed as: X t =(x t,1 ,x t,2 ,...,x t,F ), where X t is the driver feature at time step t, t is the time step, F is the dimension of the feature, x t,F is the Fth eigenvalue in this eigenvector, which represents the value of a specific feature of the driver at time step t. Applying the convolution operation on the time series, setting the size of the convolution kernel to K, the output of the convolution operation is expressed as: Among them, Z tis the output of the convolutional layer at time step t, W i is the i-th parameter of the convolution kernel, X t-i is the eigenvalue of the input sequence at time step ti;
[0022] Step C203: Capture the dynamic features of the time series through convolution. For the t-th time step, the output of TCN is expressed as: Among them, W k is the weight of the kth convolution kernel, K is the size of the convolution window, and Y t represents the output of TCN at time step t, b is the bias term, and Y t-k Represents a weighted summation of the data of past time steps;
[0023] Step C204: Apply activation function ReLU to the output of TCN to introduce nonlinearity After processing through multiple TCN layers, the final output is represented as: Among them, Y t represents the output of TCN at time step t, is the final output, is the predicted number of orders at time step T;
[0024] Step C3, model training: Divide the segmented input and output samples into training, validation, and test sets, standardize the input feature data to improve the training effect, train the CTCN model using the training set, and monitor the model performance on the validation set to prevent overfitting.
[0025] In a preferred embodiment, in step S4, the trained CTCN network model is used for prediction, daily feature data is stored in a data warehouse, and feature data of target drivers is extracted and input into the CTCN network model to predict the number of orders completed the next day. The specific steps are as follows:
[0026] Step D1: Store daily feature data in the data warehouse. Based on the target driver's ID, extract the driver's historical feature data from the data warehouse to form a time series. The extracted time series data is then divided into 30-day time windows, each of which serves as the input sequence.
[0027] Step D2: Load the trained CTCN network model from the storage location, perform a normalization operation on the sorted input sequence to ensure that the range of the feature data is consistent with the training data, and input it into the CTCN network model for prediction. The model will generate a prediction of the next day's order volume based on the past data;
[0028] Step D3: Extract the next day's order completion forecast from the model output, including the forecast date, driver ID, and forecast order completion information. Save the forecast results to the data warehouse for subsequent analysis and business decision-making.
[0029] The present application also provides a device for predicting the number of completed online ride-hailing orders based on CTCN, which specifically includes a data acquisition module, a feature storage module, a model training module, and an application prediction module;
[0030] Data collection module: extracts order data from the past 180 days from the database, performs data preprocessing, and stores it in the data warehouse;
[0031] Feature storage module: This module extracts driver characteristics, environmental characteristics, and activity incentive characteristics from the data warehouse, integrates the extracted characteristics into the "driver-day" dimension, and constructs a driver time series.
[0032] Model training module: This module segments the driver's time series, uses the segmented time series as input, and outputs the next-day completed orders in the series. It then constructs and trains the CTCN network model.
[0033] Application prediction module: Use the trained CTCN network model for prediction, store daily feature data in the data warehouse, extract the feature data of the target driver, input it into the CTCN network model, and predict the number of orders completed the next day.
[0034] The beneficial effects of the present invention are: compared with other time series prediction models, this solution uses a network structure combining CNN and TCN, the calculation process can be parallelized, and the calculation speed is faster while ensuring accuracy. The prediction is based on the time series performance of individual drivers, so that the predicted value is more in line with the individual situation of the driver. The time series prediction algorithm is used to construct a time series for each driver, and the prediction result is biased towards the individual, solving the problem of inaccurate regression prediction caused by data sparsity. BRIEF DESCRIPTION OF THE DRAWINGS
[0035] Figure 1 is a flow chart of the method of the present invention;
[0036] Figure 2 It is a structural block diagram of the present invention. DETAILED DESCRIPTION
[0037] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making creative efforts are within the scope of protection of this application.
[0038] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of the technical features indicated. Therefore, a feature specified as "first" or "second" may explicitly or implicitly include one or more of the described features. In the description of this application, "plurality" means two or more, unless otherwise specifically specified.
[0039] In the description of this application, the term "for example" is used to mean "used as an example, illustration or explanation". Any embodiment described as "for example" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is given to enable any person skilled in the art to implement and use the present invention. In the following description, details are listed for the purpose of explanation. It should be understood that a person of ordinary skill in the art will recognize that the present invention can be implemented without using these specific details. In other examples, well-known structures and processes will not be elaborated in detail to avoid obscuring the description of the present invention with unnecessary details. Therefore, the present invention is not intended to be limited to the embodiments shown, but is consistent with the widest scope consistent with the principles and features disclosed in this application.
[0040] Example 1
[0041] This embodiment provides Figure 1 A method for predicting the number of completed ride-hailing orders based on CTCN is shown, which specifically includes the following steps:
[0042] Step S1: extract the order data of the past 180 days from the database, perform data preprocessing and store it in the data warehouse;
[0043] Step S2: extracting driver characteristics, environmental characteristics, and activity incentive characteristics from the data warehouse, and integrating the extracted characteristics in the "driver-day" dimension to construct a driver time series;
[0044] Step S3: Segment the driver's time series, use the segmented time series as input, and the next-day completed orders in the series as output, build a CTCN network model, and train the CTCN network model;
[0045] Step S4: Use the trained CTCN network model to make predictions, store the daily feature data in the data warehouse, extract the feature data of the target driver, input it into the CTCN network model, and predict the number of orders completed the next day.
[0046] Preferably, in step S1, order data for the past 180 days is extracted from the database, and data preprocessing is performed to exclude dirty data, test data, and erroneous data to ensure the authenticity and accuracy of the data, and the data is stored in the data warehouse. The specific steps are as follows:
[0047] Step A1, Data Extraction: Establish a connection with the source database and extract order data from the past 180 days, including order ID, driver ID, order date, and status information;
[0048] Step A2, Data Storage: Check the extracted order data, handle missing values and outliers, convert date fields to a standard date format, ensure that the data types of all fields meet requirements, create a new table in the data warehouse to store the processed order data, and load the cleaned and processed data into the target table to ensure data integrity and consistency.
[0049] Preferably, in step S2, driver characteristics, environmental characteristics, and activity incentive characteristics are extracted from the data warehouse, and the extracted characteristics are integrated in the "driver-day" dimension to construct a time series of drivers, capture the dynamic relationship between characteristics over time, and enhance the model's learning ability for time series data. The specific steps are as follows:
[0050] Step B1, feature extraction: connect to the data warehouse, query and extract feature data related to drivers, environment and activity incentives. The driver characteristics include each driver's daily records, including whether the vehicle is effectively dispatched on that day, the dispatch time, and the number of completed orders. It also summarizes the statistical data of the past 7 days and 30 days, including the number of effective dispatch days, the total dispatch time and the number of completed orders. Among them, effective dispatch means that the driver's dispatch time is greater than 30 minutes, which is considered to be the driver's effective dispatch on that day; the environmental characteristics include the number of dispatched drivers in the city on that day, the average number of dispatched drivers per month, the average number of completed orders, as well as the daily order-to-order ratio, order matching rate, response rate and order-to-complete rate of that day. Among them, the order-to-order ratio is the number of dispatched drivers in the city on that day, the average number of dispatched drivers per month, the average number of completed orders, and the daily order-to-order ratio, order matching rate, response rate and order-to-complete rate of that day. The ratio is the number of sub-orders / the number of main orders, indicating the number of online ride-hailing tenants selected by passengers in the target city. The order matching rate is the number of orders for which drivers are found in the system / the number of sub-orders after the order is placed, reflecting the supply and demand of the order. The order response rate is the number of orders answered / the number of orders matched to drivers, reflecting the system's order scheduling tendency. The order completion rate is the number of completed orders / the number of answered orders. The activity incentive features include the driver's activity level and reward amount for the day. If there is no activity, the level is 0 and the reward is 0. The activity level, reward amount, number of orders to be completed, and the increase in the level reward are predicted for the next day based on the number of completed orders on the day.
[0051] Step B2, Feature Integration: Create a data table structure with a "driver-day" dimension for each driver to ensure that each driver has a corresponding record for each day. The extracted driver characteristics, environmental characteristics, and activity incentive characteristics are integrated according to the "driver-day" dimension to ensure that each driver's feature data for each day is correctly associated. The "driver-day" dimension data table structure integrates the extracted driver characteristics by date to form a daily feature record for each driver. The extracted environmental characteristic data and activity incentive data are integrated into the driver's feature record by date and city.
[0052] Step B3, construct time series: using "driver-day" as the dimension, arrange the integrated feature data in chronological order and construct a time series data table X = (x1, x2, ..., x n ), where x n It is the feature vector of the nth day. Each row represents the complete feature record of a driver on a certain day. The constructed time series data is saved in the data warehouse for subsequent analysis and model training.
[0053] Preferably, in step S3, the driver's time series is segmented, the segmented time series is used as input, and the next-day completed orders in the series are used as output. A CTCN network model is constructed and trained. The CTCN network model combines CNN and TCN, and can simultaneously extract spatial features and temporal features, thereby enhancing the understanding of complex time series data. The specific steps are as follows:
[0054] Step C1: Split the time series: Extract the time series feature data of drivers from the data warehouse, traverse the entire time series, and split it into multiple segments at 30-day intervals, with each segment containing 30 days of data. For each segment, the feature data in that segment is used as input, and the output is the number of orders completed the next day on the last day of the segment. Each segment can generate new input and output pairs, increasing the amount of training data for the model and improving its generalization ability.
[0055] Step C2: Constructing a CTCN network model: The CTCN network model is a new network architecture that combines the CNN network and the TCN network. CNN is a convolutional neural network used to extract features and model spatial relationships in data, and TCN is a temporal convolutional network used to capture dynamic changes in time series data. The model further includes the following steps:
[0056] Step C201: Define the model structure: The input layer accepts 30 days of time series data, uses a one-dimensional convolutional layer to extract spatial features, adds a temporal convolutional layer to capture the dynamic changes of the time series, uses a pooling layer to reduce the dimensionality of the data, uses a fully connected layer to process the features extracted by the convolutional layer, maps them to the final prediction result, and finally, the output layer outputs the prediction result of the number of orders completed for the next day;
[0057] Step C202, convolution operation: the feature vector of each driver at different time steps is expressed as: X t =(x t,1 ,x t,2 ,...,x t,F ), where X t is the driver feature at time step t, t is the time step, F is the dimension of the feature, x t,F is the Fth eigenvalue in this eigenvector, which represents the value of a specific feature of the driver at time step t. Applying the convolution operation on the time series, setting the size of the convolution kernel to K, the output of the convolution operation is expressed as: Among them, Z t is the output of the convolutional layer at time step t, W i is the i-th parameter of the convolution kernel, X t-i is the eigenvalue of the input sequence at time step ti;
[0058] Step C203: Capture the dynamic features of the time series through convolution. For the t-th time step, the output of TCN is expressed as: Among them, W k is the weight of the kth convolution kernel, K is the size of the convolution window, and Y t represents the output of TCN at time step t, b is the bias term, and Y t-k Represents a weighted summation of the data of past time steps;
[0059] Step C204: Apply activation function ReLU to the output of TCN to introduce nonlinearity After processing through multiple TCN layers, the final output is represented as: Among them, Y t represents the output of TCN at time step t, is the final output, is the predicted number of orders at time step T;
[0060] Step C3, model training: Divide the split input and output samples into training sets, validation sets, and test sets. Standardize the input feature data to eliminate dimensional differences between features and ensure that all features are in the same range. This helps speed up model convergence and improve training results. Use the training set to train the CTCN model, while monitoring the model performance on the validation set. Set an early stopping mechanism to stop training when the performance of the validation set no longer improves to avoid overfitting of the model on the training set.
[0061] Preferably, in step S4, the trained CTCN network model is used for prediction, which can fully utilize the features and patterns captured by historical data to enhance the accuracy of the prediction of the next day's order completion volume. The daily feature data is stored in the data warehouse, and the feature data of the target driver is extracted and input into the CTCN network model to predict the next day's order completion volume. The specific steps are as follows:
[0062] Step D1: Store daily feature data in the data warehouse so that each prediction is based on the latest and most relevant data, ensuring timeliness and accuracy. Based on the target driver's ID, extract the driver's historical feature data from the data warehouse to form a time series. This extracted time series data is then divided into 30-day time windows, each of which serves as the input sequence.
[0063] Step D2: Load the trained CTCN network model from its storage location, normalize the organized input sequence to ensure that the feature data range is consistent with the training data, and then input it into the CTCN network model for prediction. The model will generate a forecast of the next day's order completion volume based on past data. The CTCN model can effectively integrate these influencing factors by combining driver characteristics, environmental characteristics, and other activity incentives to improve the model's understanding of order completion volume.
[0064] Step D3: Extract the next day's order completion forecast from the model output, including the forecast date, driver ID, and forecast order completion information. Save the forecast results to the data warehouse for subsequent analysis and business decision-making.
[0065] Example 2
[0066] This embodiment provides a device for predicting the number of completed ride-hailing orders based on CTCN, which specifically includes a data acquisition module, a feature storage module, a model training module, and an application prediction module.
[0067] Data collection module: extracts order data from the past 180 days from the database, performs data preprocessing, and stores it in the data warehouse;
[0068] Feature storage module: This module extracts driver characteristics, environmental characteristics, and activity incentive characteristics from the data warehouse, integrates the extracted characteristics into the "driver-day" dimension, and constructs a driver time series.
[0069] Model training module: This module segments the driver's time series, uses the segmented time series as input, and outputs the next-day completed orders in the series. It then constructs and trains the CTCN network model.
[0070] Application prediction module: Use the trained CTCN network model for prediction, store daily feature data in the data warehouse, extract the feature data of the target driver, input it into the CTCN network model, and predict the number of orders completed the next day.
[0071] It should be noted that, in the above embodiments, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.
[0072] It will be understood by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0073] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0074] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.
[0075] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.
[0076] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.
[0077] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.
Claims
1. A method for predicting the number of completed ride-hailing orders based on CTCN, characterized by: The specific steps include: Step S1: extract the order data of the past 180 days from the database, perform data preprocessing and store it in the data warehouse; Step S2: Extract driver characteristics, environmental characteristics, and activity incentive characteristics from the data warehouse, and integrate the extracted characteristics into the "driver-day" dimension to construct a driver time series; Step S3: Segment the driver's time series, use the segmented time series as input, and the next-day completed orders in the series as output, build a CTCN network model, and train the CTCN network model; Step S4: Use the trained CTCN network model to make predictions, store the daily feature data in the data warehouse, extract the feature data of the target driver, input it into the CTCN network model, and predict the number of orders completed the next day.
2. The method for predicting the number of completed ride-hailing orders based on CTCN according to claim 1 is characterized by: In step S1, the order data for the past 180 days is extracted from the database, and the data is pre-processed and stored in the data warehouse. The specific steps are as follows: Step A1, Data Extraction: Establish a connection with the source database and extract order data from the past 180 days, including order ID, driver ID, order date, and status information; Step A2, Data Storage: Check the extracted order data, handle missing values and outliers, convert date fields to a standard date format, create a new table in the data warehouse to store the processed order data, and load the cleaned and processed data into the target table.
3. The method for predicting the number of completed ride-hailing orders based on CTCN according to claim 1 is characterized by: In step S2, driver characteristics, environmental characteristics, and activity incentive characteristics are extracted from the data warehouse, and the extracted characteristics are integrated in the "driver-day" dimension to construct a driver time series. The specific steps are as follows: Step B1, Feature Extraction: Connect to the data warehouse to query and extract feature data related to drivers, environment, and activity incentives. The driver features include each driver's daily records, including whether the driver was effectively dispatched that day, the duration of the dispatch, and the number of orders completed. Statistical data from the past 7 days and 30 days is summarized, including the number of effective dispatch days, total dispatch duration, and number of orders completed. The environmental features include the number of dispatched drivers in the city on that day, the average monthly number of dispatched drivers, the average number of orders completed, as well as the daily order-to-order ratio, order matching rate, response rate, and order completion rate. The activity incentive features include the driver's activity tier and reward amount for that day, and based on the number of orders completed that day, the next day's activity tier, reward amount, number of orders to be completed, and the increase in tier rewards are predicted. Step B2, Feature Integration: Create a data table structure with the "Driver-Day" dimension for each driver, ensuring that each driver has a corresponding record every day. Integrate the extracted driver characteristics, environmental characteristics, and activity incentive characteristics according to the "Driver-Day" dimension. Step B3, construct time series: Using "driver-day" as the dimension, arrange the integrated feature data in chronological order and construct a time series data table X = (x1, x2, ..., x n ), where x n It is the feature vector of the nth day. Each row represents the complete feature record of a driver on a certain day. The constructed time series data is saved in the data warehouse for subsequent model training.
4. The method for predicting the number of completed ride-hailing orders based on CTCN according to claim 1, characterized in that: In step S3, the driver's time series is segmented, the segmented time series is used as input, and the next day's completed orders in the series are used as output. The CTCN network model is constructed and trained. The specific steps are as follows: Step C1: Split the time series: Extract the time series feature data of drivers from the data warehouse, traverse the entire time series, and split it into multiple segments at 30-day intervals, with each segment containing 30 days of data. For each segment, use the feature data in the segment as input, and output the number of completed orders the next day after the last day of the segment; Step C2: Construct a CTCN network model: The CTCN network model is a new network architecture that combines the CNN network and the TCN network. CNN is a convolutional neural network used to extract features and model spatial relationships in data, while TCN is a temporal convolutional network used to capture dynamic changes in time series data. Step C3, model training: Divide the segmented input and output samples into training, validation, and test sets, standardize the input feature data, train the CTCN model using the training set, and monitor the model performance on the validation set to prevent overfitting.
5. The method for predicting the number of completed ride-hailing orders based on CTCN according to claim 4 is characterized by: In the step C2 of constructing the CTCN network model, the CTCN network model is a new network architecture that combines the CNN network and the TCN network, and further includes the following steps: Step C201: Define the model structure: The input layer accepts 30 days of time series data, uses a convolutional layer to extract spatial features, a temporal convolutional layer to capture the dynamic changes of the time series, uses a pooling layer to reduce the dimensionality of the data, uses a fully connected layer to process the features extracted by the convolutional layer, maps them to the final prediction result, and finally, the output layer outputs the prediction result of the number of orders completed the next day; Step C202, convolution operation: the feature vector of each driver at different time steps is expressed as: X t =(x t,1 ,x t,2 ,...,x t,F ), where X t is the driver feature at time step t, t is the time step, F is the dimension of the feature, x t,F is the Fth eigenvalue in this eigenvector. Applying the convolution operation on the time series, setting the size of the convolution kernel to K, the output of the convolution operation is expressed as: Among them, Z t is the output of the convolutional layer at time step t, W i is the i-th parameter of the convolution kernel, X t-i is the eigenvalue of the input sequence at time step ti; Step C203: Capture the dynamic features of the time series through convolution. For the t-th time step, the output of TCN is expressed as: Among them, W k is the weight of the kth convolution kernel, K is the size of the convolution window, and Y t represents the output of TCN at time step t, b is the bias term, and Y t-k Represents a weighted summation of the data of past time steps; Step C204: Apply activation function ReLU to the output of TCN to introduce nonlinearity After processing through multiple TCN layers, the final output is represented as: Among them, Y t represents the output of TCN at time step t, is the final output, is the predicted order quantity at time step T.
6. The method for predicting the number of completed ride-hailing orders based on CTCN according to claim 1, characterized in that: In step S4, the trained CTCN network model is used for prediction. The daily feature data is stored in the data warehouse, and the feature data of the target driver is extracted and input into the CTCN network model to predict the number of orders completed the next day. The specific steps are as follows: Step D1: Store daily feature data in the data warehouse. Based on the target driver's ID, extract the driver's historical feature data from the data warehouse to form a time series. The extracted time series data is then divided into 30-day time windows, each of which serves as the input sequence. Step D2: Load the trained CTCN network model from the storage location, perform a normalization operation on the sorted input sequence to ensure that the range of the feature data is consistent with the training data, and input it into the CTCN network model for prediction. The model will generate a prediction of the next day's order volume based on the past data; Step D3: Extract the next day's order completion forecast from the model output, including the forecast date, driver ID, and forecast order completion information. Save the forecast results to the data warehouse for subsequent analysis and business decision-making.
7. A device for predicting the number of completed online ride-hailing orders based on CTCN is applied to a method for predicting the number of completed online ride-hailing orders based on CTCN as described in any one of claims 1-6, characterized in that: Includes data acquisition module, feature storage module, model training module, and application prediction module; Data collection module: extracts order data from the past 180 days from the database, performs data preprocessing, and stores it in the data warehouse; Feature storage module: This module extracts driver characteristics, environmental characteristics, and activity incentive characteristics from the data warehouse and integrates the extracted characteristics into the "driver-day" dimension to construct a driver time series. Model training module: This module segments the driver's time series, uses the segmented time series as input, and outputs the next-day completed orders in the series. It then constructs and trains the CTCN network model. Application prediction module: Use the trained CTCN network model for prediction, store daily feature data in the data warehouse, extract the feature data of the target driver, input it into the CTCN network model, and predict the number of orders completed the next day.