A method and system for allocating online car-hailing orders

Abstract

The present invention provides a method and system for allocating online car-hailing orders, and relates to the technical field of order management. The method comprises: after receiving an order to be allocated, determining order itinerary information and a vehicle screening area according to the order to be allocated; obtaining order completion data of multiple online car-hailing vehicles in an idle state on the day and calculating a fatigue index; if there is no online car-hailing vehicle with a fatigue index less than a preset health index threshold, constructing a target vehicle set, collecting historical order data and target return locations of each online car-hailing vehicle in the target vehicle set; calculating the order discrete index of each online car-hailing vehicle in the target vehicle set, selecting a target prediction model to perform a prediction order analysis on each online car-hailing vehicle in the target vehicle set, and generating the order confidence of each online car-hailing vehicle; calculating the return index of each online car-hailing vehicle, and selecting the online car-hailing vehicle with the smallest return index to allocate the order to be allocated. The present invention achieves the improvement of the efficiency of order allocation and the work quality of drivers.

Description

Technical Field

[0001] The present invention relates to the technical field of order management, and in particular to a method and system for allocating online car-hailing orders. Background Art

[0002] Order allocation is a crucial link in the online car-hailing industry, which directly affects the user experience and the driver's work efficiency. However, the current order allocation methods of some online car-hailing companies have certain limitations. Traditional order allocation methods are usually simply determined based on the distance between the passenger and the driver, without considering the driver's driving fatigue and the round-trip distance involved when the driver chooses to go home and rest after completing the order if he does not want to continue to take orders. Long driving will increase the driver's fatigue, affecting his driving safety and service quality. If the driver chooses to go home and rest after completing the order, if the journey home is long, it will further increase the driver's burden and fatigue. Summary of the invention

[0003] In order to solve the above technical problems, the present invention proposes a method and system for allocating online car-hailing orders, which comprehensively considers the driver's driving fatigue level and the round-trip distance involved when the driver does not want to continue to take orders after completing an order and chooses to go home to rest, and selects a more suitable vehicle for order allocation, thereby effectively improving the efficiency of order allocation and the driver's work quality and enhancing the user experience.

[0004] To achieve the above-mentioned purpose, the first aspect of the present invention provides a method for allocating online car-hailing orders, comprising:

[0005] After receiving the order to be allocated, determine the order itinerary information according to the order to be allocated, including the starting location of the itinerary, the ending location of the itinerary and the estimated duration of the itinerary, and determine the vehicle screening area based on the order itinerary information;

[0006] Obtain the order completion data of multiple online-hailing vehicles in the vehicle screening area that are idle on the same day, and calculate the fatigue index of each online-hailing vehicle based on the order completion data;

[0007] Determine whether there is an online-hailing vehicle whose fatigue index is less than a preset health index threshold among multiple online-hailing vehicles in an idle state in a vehicle screening area;

[0008] If there is no online-hailing vehicle with a fatigue index less than a preset health index threshold among the multiple online-hailing vehicles in the vehicle screening area that are idle, a target vehicle set is constructed based on multiple online-hailing vehicles in the vehicle screening area that are idle and have fatigue indexes not less than the preset health index threshold, and historical order data and target return locations of each online-hailing vehicle in the target vehicle set are collected;

[0009] Based on the historical order data, the order dispersion index of each online-hailing vehicle in the target vehicle set is calculated. According to the order dispersion index, the target prediction model is selected to perform a prediction order analysis on each online-hailing vehicle in the target vehicle set, and the order confidence of each online-hailing vehicle is generated.

[0010] The return index of each online-hailing vehicle is calculated based on the order acceptance confidence and the target return location, and the online-hailing vehicle with the smallest return index is selected to allocate the orders to be allocated.

[0011] Preferably, the fatigue index of each online car-hailing vehicle is calculated based on the order completion data, including:

[0012] The total mileage, total number of orders, average completion time and total rest time corresponding to the order completion data of each online car-hailing vehicle are determined based on the order completion data, and the fatigue index is calculated using the following formula:

[0013] F=w1C1+w2C2+w3C3+w4C4

[0014] Where F is the fatigue index of the online car-hailing service, C1 is the total mileage of passengers dropped off, C2 is the total number of orders, C3 is the average completion time, C4 is the total rest time, and w1, w2, w3 and w4 are the weight parameters of the total mileage of passengers dropped off, the total number of orders, the average completion time and the total rest time, respectively.

[0015] Preferably, in the process of determining whether there is an online-hailing vehicle whose fatigue index is less than a preset health index threshold among the plurality of online-hailing vehicles in an idle state in the vehicle screening area, the method further includes:

[0016] If there is an online-hailing vehicle with a fatigue index lower than the preset health index threshold among multiple idle online-hailing vehicles in the vehicle screening area and the number is 1, the order to be allocated will be allocated to the online-hailing vehicle with a fatigue index lower than the preset health index threshold; if there is and the number is not 1, the online-hailing vehicle closest to the starting location of the trip will be selected for order allocation based on the order itinerary information.

[0017] Preferably, the order discrete index of each online-hailing vehicle in the target vehicle set is calculated based on the historical order data, including:

[0018] The total number of orders, order time span parameters, order frequency and order distribution parameters of each online car-hailing vehicle are determined based on historical order data, and the order discrete index of each online car-hailing vehicle is calculated using the following formula:

[0019] D=u1m1+u2m2+u3m3+u4m4

[0020] Where D is the order dispersion index, m1 is the total number of orders, m2 is the order time span parameter, m3 is the order frequency, m4 is the order distribution parameter, u1, u2, u3 and u4 are the weight parameters of the total number of orders, order time span parameter, order frequency and order distribution parameter respectively;

[0021] Among them, the order time span parameter is the number of days between the first and last orders in the historical order data, and the order distribution parameter is the standard deviation of the orders in the historical order data. The calculation formula is as follows:

[0022]

[0023] Where n is the total number of days corresponding to the historical order data, x i is the number of orders on the i-th day, is the average daily number of orders in the historical order data.

[0024] Preferably, a target prediction model is selected according to the order dispersion index to perform a prediction order analysis on each online-hailing vehicle in the target vehicle set, and the order confidence of each online-hailing vehicle is generated, including:

[0025] If the order discrete index of the online car-hailing vehicle is greater than the preset discrete threshold, the personalized prediction model corresponding to the online car-hailing vehicle is selected as the target prediction model, and the target order data of the online car-hailing vehicle within the preset reference range is input into the target prediction model to generate the order confidence of the online car-hailing vehicle;

[0026] If the order discrete index of the online car-hailing service is not greater than the preset discrete threshold, the general prediction model is selected as the target prediction model, and the target order data of the online car-hailing service within the preset reference range is input into the target prediction model to generate the order confidence of the online car-hailing service.

[0027] Preferably, the personalized prediction model and the general prediction model further include:

[0028] Obtain historical order data of multiple online ride-hailing vehicles, extract the second order time series data of each set of historical order data and construct a general training data set, and train a general prediction model based on the general training data set;

[0029] Based on multiple groups of historical order data, the order discrete index of each online-hailing vehicle is calculated, and multiple groups of historical order data with order discrete index greater than a preset discrete threshold are screened out. The second order receiving time series data of each group of historical order data is extracted and a personalized training data set for each group of historical order data is constructed. Based on the multiple groups of personalized training data sets, personalized prediction models for multiple online-hailing vehicles are trained respectively.

[0030] Preferably, the return index of each online car-hailing vehicle is calculated based on the order confidence and the target return location, including:

[0031] The return index is calculated using the following formula: R = (1-P)L, where R is the return index, P is the order confidence, and L is the return mileage, where the return mileage is the driving distance between the end location of the trip and the target return location.

[0032] A second aspect of the present invention provides an online car-hailing order allocation system, which is used to implement the above-mentioned online car-hailing order allocation method, including:

[0033] An information extraction module is used to determine the order itinerary information according to the order to be allocated after receiving the order to be allocated, including the starting location of the trip, the ending location of the trip and the estimated trip duration, and determine the vehicle screening area based on the order itinerary information;

[0034] The fatigue analysis module is used to obtain the order completion data of multiple online-hailing vehicles in the vehicle screening area that are idle on the same day, and calculate the fatigue index of each online-hailing vehicle based on the order completion data;

[0035] The vehicle analysis module is used to determine whether there is an online-hailing vehicle with a fatigue index less than a preset health index threshold among the multiple online-hailing vehicles in the vehicle screening area that are idle, and when there is no online-hailing vehicle with a fatigue index less than the preset health index threshold among the multiple online-hailing vehicles in the vehicle screening area that are idle, a target vehicle set is constructed based on the multiple online-hailing vehicles in the vehicle screening area that are idle and have fatigue indexes not less than the preset health index threshold, and historical order data and target return locations of each online-hailing vehicle in the target vehicle set are collected;

[0036] The order analysis module is used to calculate the order dispersion index of each online-hailing vehicle in the target vehicle set based on historical order data;

[0037] The order acceptance analysis module is used to select the target prediction model according to the order dispersion index to perform order prediction analysis on each online-hailing vehicle in the target vehicle set and generate the order acceptance confidence of each online-hailing vehicle;

[0038] The order allocation module is used to calculate the return index of each online-hailing vehicle based on the order acceptance confidence and the target return location, and select the online-hailing vehicle with the smallest return index to allocate the orders to be allocated.

[0039] Preferably, the order allocation module further includes:

[0040] If there is an online-hailing vehicle with a fatigue index lower than the preset health index threshold among multiple idle online-hailing vehicles in the vehicle screening area and the number is 1, the order to be allocated will be allocated to the online-hailing vehicle with a fatigue index lower than the preset health index threshold; if there is and the number is not 1, the online-hailing vehicle closest to the starting location of the trip will be selected for order allocation based on the order itinerary information.

[0041] The present invention has the following beneficial effects:

[0042] The present invention comprehensively considers the driver's driving fatigue level and the round-trip distance involved when the driver chooses to go home and rest after completing an order if he does not want to continue accepting orders. Specifically, the fatigue index is calculated to characterize the driver's fatigue level, and a prediction model suitable for predicting the order needs of different drivers is selected through the order discrete index to obtain a more accurate prediction result. The return index of the online car-hailing service is calculated to quantify the impact of the round-trip distance involved when the driver chooses to go home and rest after completing an order if he does not want to continue accepting orders. Consideration is taken from multiple aspects to select suitable vehicles for different orders for allocation, thereby effectively improving the efficiency of order allocation and the work quality of drivers and enhancing the user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 The present invention is a flowchart of a method for allocating online car-hailing orders in an embodiment of the present invention. DETAILED DESCRIPTION

[0044] In order to make those skilled in the art better understand the technical solutions in the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0045] See also Figure 1 , a method for allocating online car-hailing orders provided in an embodiment of the present invention specifically includes the following steps:

[0046] S10. After receiving the order to be allocated, determine the order itinerary information according to the order to be allocated, and determine the vehicle screening area based on the order itinerary information.

[0047] In this embodiment, the order itinerary information includes at least information such as the starting location of the trip, the ending location of the trip, and the estimated duration of the trip. Users who need a taxi can send a request through a smart mobile device. The online car-hailing platform generates a corresponding order to be assigned based on the user's taxi demand, and at the same time preliminarily selects an online car-hailing platform for the starting location of the trip involved in the order. Among them, the vehicle screening area can be determined according to the starting location of the trip, such as a circular area or a rectangular area centered on the starting location of the trip, which is not specifically limited in this embodiment.

[0048] S20. Obtain order completion data for a plurality of online-hailing vehicles in an idle state in a vehicle screening area on the day, and calculate a fatigue index of each online-hailing vehicle based on the order completion data.

[0049] It is worth noting that for multiple idle online-hailing vehicles in the vehicle screening area, the existing technology of directly allocating orders based on distance has certain limitations. In this embodiment, for idle online-hailing vehicles in a certain area, the order completion data of these online-hailing drivers on the same day are first analyzed to analyze the driver's fatigue index. It should be added that in this embodiment, for the convenience of description, the online-hailing vehicle is temporarily used to replace the driver driving the online-hailing vehicle, that is, the fatigue index of the online-hailing vehicle actually represents the fatigue level of the driver driving the online-hailing vehicle.

[0050] S30. Determine whether there is an online-hailing vehicle whose fatigue index is less than a preset health index threshold among the plurality of online-hailing vehicles in an idle state in the vehicle screening area.

[0051] S40. If it does not exist, a target vehicle set is constructed based on a plurality of online-hailing vehicles in the vehicle screening area that are idle and whose fatigue index is not less than a preset health index threshold, and historical order data and target return locations of each online-hailing vehicle in the target vehicle set are collected.

[0052] In this embodiment, there is no online-hailing vehicle with a fatigue index lower than a preset health index threshold among multiple idle online-hailing vehicles in the vehicle screening area, and considering that some drivers may choose to stop accepting orders and go home to rest after completing a small number of orders or a long total mileage of dropping off passengers on the day, for these online-hailing vehicles, the historical order data and target return locations corresponding to these online-hailing vehicles are extracted, and the matching degree between different online-hailing vehicles and the orders to be assigned is further analyzed based on this information.

[0053] S50. Calculate the order discrete index of each online-hailing vehicle in the target vehicle set based on the historical order data, select a target prediction model according to the order discrete index to perform a prediction order analysis on each online-hailing vehicle in the target vehicle set, and generate an order confidence level for each online-hailing vehicle.

[0054] In this embodiment, the order dispersion index is used to characterize the reference value of historical order data. Some historical order data of online ride-hailing vehicles are analyzed through a pre-trained deep learning model to predict the possibility that online ride-hailing drivers will choose to continue accepting orders in the future, and the corresponding order confidence of the online ride-hailing vehicle in a certain period of time in the future is obtained, that is, the probability of continuing to accept orders.

[0055] S60. Calculate the return index of each online-hailing vehicle based on the order acceptance confidence and the target return location, and select the online-hailing vehicle with the smallest return index to allocate the order to be allocated.

[0056] In this embodiment, the role of the order acceptance confidence is to characterize the probability that the online car-hailing vehicle will choose to continue accepting orders after completing the currently assigned order. It can also be regarded as the probability of choosing to stop accepting orders and go home to rest. The mileage corresponding to the return trip of the online car-hailing vehicle is determined by comprehensively considering the target return location and the end location of the trip involved in the order to be assigned, thereby calculating the return index of each online car-hailing vehicle.

[0057] In this embodiment, the return index of each online-hailing vehicle is calculated based on the order confidence and the target return location, including:

[0058] The return index is calculated using the following formula: R = (1-P)L, where R is the return index, P is the order confidence, and L is the return mileage, where the return mileage is the driving distance between the end location of the trip and the target return location.

[0059] It is worth noting that the greater the confidence level of the order generated by the model prediction and the smaller the return mileage, the more likely the online car-hailing driver is to continue accepting orders after completing the order to be assigned. Even if he chooses not to continue accepting orders but returns to rest, the distance traveled is shorter. Therefore, considering the fatigue level of different online car-hailing drivers, selecting the online car-hailing driver with the smallest return index to allocate the order to be assigned can greatly reduce the burden and fatigue of the online car-hailing driver and effectively improve the efficiency of order allocation and the quality of driver's work.

[0060] It is worth explaining that the above-mentioned allocation of pending orders specifically refers to giving priority to sending pending orders to eligible online ride-hailing vehicles. After receiving relevant order information, the driver can choose whether to accept the order according to actual needs. It is not limited to mandatory allocation of orders, but is only used to represent the priority of information sending, that is, the order is first provided to more suitable online ride-hailing drivers for selection, so as to avoid excessive fatigue of the drivers and provide protection for the drivers' work safety.

[0061] As an optional implementation scheme, in the above step S20, the fatigue index of each online-hailing vehicle is calculated based on the order completion data, specifically including:

[0062] The total mileage, total number of orders, average completion time and total rest time corresponding to the order completion data of each online car-hailing vehicle are determined based on the order completion data. Among them, the total mileage can directly reflect the driver's driving load. Long-distance driving can cause fatigue; the number of completed orders can reflect factors such as the driver's frequent boarding and alighting, and interaction with passengers, which will also affect the degree of fatigue. More orders may mean more processing of starting and ending locations, increasing the workload; the average completion time reflects the average time required for the driver to complete an order. If the average order time is long, it may mean that the driver needs to drive continuously for a long time, increasing the risk of fatigue; the total rest time can more comprehensively represent the driver's rest during the entire working period. Compared with considering a certain period of rest alone, this method can reflect whether the driver has received enough rest to relieve fatigue. A longer total rest time means that the driver has enough time to restore his energy and physical strength.

[0063] In this embodiment, the data for calculating the fatigue index is specifically the order completion data of the online car-hailing service on that day, which is used to quantify the driver's fatigue level on that day, and the fatigue index is calculated using the following formula:

[0064] F=w1C1+w2C2+w3C3+w4C4

[0065] In the formula, F is the fatigue index of the online car-hailing service, C1 is the total mileage of passengers dropped off, C2 is the total number of orders, C3 is the average completion time, C4 is the total rest time, w1, w2, w3 and w4 are the weight parameters of the total mileage of passengers dropped off, the total number of orders, the average completion time and the total rest time, respectively. The weights are set according to the actual situation to reflect the relative importance of each factor to fatigue.

[0066] The fatigue index calculated by the above formula can accurately reflect the driver's fatigue level, thereby better protecting the safety of drivers and passengers during the order allocation process and improving the service quality of online ride-hailing services.

[0067] As an optional implementation scheme, for step S30, in the process of determining whether there is an online-hailing vehicle whose fatigue index is less than a preset health index threshold among the plurality of online-hailing vehicles in an idle state in the vehicle screening area, the process further includes:

[0068] If there is an online-hailing vehicle with a fatigue index lower than the preset health index threshold among multiple idle online-hailing vehicles in the vehicle screening area and the number is 1, the order to be allocated will be allocated to the online-hailing vehicle with a fatigue index lower than the preset health index threshold; if there is and the number is not 1, the online-hailing vehicle closest to the starting location of the trip will be selected for order allocation based on the order itinerary information.

[0069] In this embodiment, the role of the preset health index threshold is to quantitatively measure whether the fatigue level of the online car-hailing driver exceeds a certain limit. The fact that the fatigue index is greater than or equal to the preset health index threshold does not mean that the online car-hailing driver is already in a state of fatigue. In actual application, a safety warning can be issued based on the preset health index threshold, which means that after the fatigue index is greater than or equal to the preset health index threshold, the driver may need to temporarily postpone accepting orders and rest in order to maintain a good mental state. The preset health index threshold can be reasonably set according to actual needs.

[0070] In this embodiment, for the order allocation process, the online car-hailing vehicle with a fatigue index less than the preset health index threshold is preferred for continued allocation, so that some online car-hailing drivers with longer working hours can have a better rest. If the number of online car-hailing vehicles in the idle state in the vehicle screening area is 1, it can be directly allocated. If the number of online car-hailing vehicles is large, it can be allocated according to the principle of proximity to reduce the waiting time of users. It should be noted that selecting the online car-hailing vehicle closest to the starting position of the trip based on the order trip information does not specifically refer to the absolute distance, but the vehicle's driving distance. In actual applications, traffic flow information of different paths can also be considered to predict the actual arrival time, and select the online car-hailing vehicle that can arrive first in theory for order allocation. When the online car-hailing driver is not in a fatigued state, the existing technology can also be used to reasonably select a suitable online car-hailing vehicle for order allocation, which is not specifically limited here.

[0071] As an optional implementation scheme, in the above step S50, the order discrete index of each online-hailing vehicle in the target vehicle set is calculated based on the historical order data, specifically including:

[0072] The total number of orders, order time span parameters, order frequency and order distribution parameters of each online car-hailing vehicle are determined based on historical order data. The total number of orders refers to the number of all completed orders involved in the historical data, reflecting the driver's total historical workload; the order time span parameter indicates the time range of the historical order data, that is, the number of days between the first and last orders in the historical order data. A longer time span may mean that the driver has more work experience and a longer service history; the order frequency refers to the average number of orders completed by the driver per day, reflecting the driver's service activity and work efficiency; the uniformity of order time period distribution reflects the distribution of orders completed by the driver every day over a period of time, reflecting the stability of the workload.

[0073] The order discrete index of each online car-hailing vehicle is calculated using the following formula:

[0074] D=u1m1+u2m2+u3m3+u4m4

[0075] Where D is the order dispersion index, m1 is the total number of orders, m2 is the order time span parameter, m3 is the order frequency, m4 is the order distribution parameter, u1, u2, u3 and u4 are the weight parameters of the total number of orders, order time span parameter, order frequency and order distribution parameter respectively.

[0076] The above order distribution parameter is specifically the standard deviation of orders in the historical order data, and the calculation formula is as follows:

[0077]

[0078] Where n is the total number of days corresponding to the historical order data, x i is the number of orders on the i-th day, is the average daily number of orders in the historical order data.

[0079] In this embodiment, the weight parameters of the total number of orders, order time span parameters, order frequency and order distribution parameters are reasonably set according to actual needs, and the calculated order discrete index can reflect the reference value of the historical order data of each online car-hailing vehicle.

[0080] It is worth noting that in the process of training a prediction model or predicting online ride-hailing orders through a prediction model, if the amount of data for each driver is sufficient, a personalized prediction model can be trained during the model training process. In this case, the model can capture the personalized characteristics of each driver and provide more accurate predictions. In the actual prediction process, the personalized model trained for each online ride-hailing driver can also be used to make more accurate order predictions. If the amount of data for some drivers is relatively small and insufficient to support the training of personalized models, in this case, starting from the overall situation, a generalized general prediction model can be obtained by training the historical data of multiple online ride-hailing drivers to capture common laws. For example, in some specific time periods, the overall order frequency of multiple online ride-hailing drivers is high or low. In the actual prediction process, for some new drivers with a small number of completed orders, it is impossible to train a personalized model due to insufficient data. In this case, a generalized general prediction model that can capture common laws can be selected for order prediction.

[0081] In this embodiment, the calculation of the order dispersion index comprehensively considers the distribution of multiple historical orders. To train and use a personalized prediction model to accurately predict the order acceptance situation of some online car-hailing drivers, conventional methods, such as simply considering the total number of orders completed by the driver, believe that after reaching a certain number of orders, the data is of reference value and can be used to train a personalized model. This method has certain limitations, because if the number of orders is concentrated in certain days instead of being evenly distributed over multiple days, it may affect the generalization ability of the model. The prediction accuracy of the trained model may have large errors in the actual prediction results because the data distribution is relatively concentrated. In order to more accurately evaluate the generalization of historical order data, the above formula is used in this embodiment to calculate the order dispersion index of each group of historical order data, which can more accurately measure whether the historical order data is of reference value, so as to improve the prediction ability and reliability of the trained personalized model.

[0082] As an optional implementation scheme, in the above step S50, a target prediction model is selected according to the order dispersion index to perform a prediction order analysis on each online-hailing vehicle in the target vehicle set to generate the order confidence of each online-hailing vehicle, specifically including:

[0083] If the order discrete index of the online car-hailing service is greater than the preset discrete threshold, the personalized prediction model corresponding to the online car-hailing service is selected as the target prediction model, and the target order data of the online car-hailing service within the preset reference range is input into the target prediction model to generate the order confidence of the online car-hailing service.

[0084] In this embodiment, the order discrete index of the online car-hailing vehicle is greater than the preset discrete threshold. The reasonable setting of the preset discrete threshold can indicate that the online car-hailing vehicle driver has relatively rich work experience, and the historical order data of the online car-hailing vehicle driver has a high reference value. A personalized prediction model that conforms to the personalized behavior of the online car-hailing vehicle driver can be trained. The personalized prediction model corresponding to the online car-hailing vehicle driver is used as the target prediction model. The target order data is selected from the historical order data of the online car-hailing vehicle according to the preset reference range and analyzed through the personalized prediction model to predict the confidence level of the online car-hailing vehicle driver in accepting orders in the future.

[0085] If the order discrete index of the online car-hailing service is not greater than the preset discrete threshold, the general prediction model is selected as the target prediction model, and the target order data of the online car-hailing service within the preset reference range is input into the target prediction model to generate the order confidence of the online car-hailing service.

[0086] In this embodiment, if the order discrete index of the online car-hailing vehicle is not greater than the preset discrete threshold, it means that the historical order data of the online car-hailing vehicle is not rich enough and is not representative. In this case, a general prediction model is used to analyze the target order data of the online car-hailing vehicle driver within the preset reference range, and the prediction results are consistent with the common laws of a large number of online car-hailing vehicles.

[0087] In this embodiment, the above-mentioned personalized prediction model and general prediction model further include:

[0088] The general prediction model is based on the historical order data of multiple online ride-hailing vehicles. The second order time series data of each set of historical order data is extracted and a general training data set is constructed. It is trained with the general training data set and captures the common characteristics of a large number of online ride-hailing drivers. It is suitable for predicting the order-taking behavior of new drivers and other online ride-hailing drivers who do not have enough work experience.

[0089] The personalized prediction model is trained based on a personalized training data set constructed based on the historical order data of each online car-hailing vehicle. Specifically, the order discrete index of each online car-hailing vehicle is calculated based on multiple sets of historical order data, and multiple sets of historical order data with order discrete index greater than a preset discrete threshold are screened out, that is, some historical order data with reference value and model training value are screened out, the second order time series data of each set of historical order data is extracted and a personalized training data set of each set of historical order data is constructed, and personalized prediction models of multiple online car-hailing vehicles are trained based on multiple sets of personalized training data sets.

[0090] In this embodiment, the training of the personalized prediction model can be carried out in advance. For example, the order discrete index of the historical order data of different online car-hailing vehicles is analyzed and calculated once a day. After determining that there is historical order data with an order discrete index greater than a preset discrete threshold, if the personalized prediction model of the online car-hailing vehicle has not been trained, the model training operation can be executed. The subsequent order prediction of the online car-hailing vehicle can be analyzed using the corresponding personalized prediction model instead of the general prediction model to improve the accuracy of the prediction.

[0091] It is worth noting that after the general prediction model is trained, the model can be updated regularly, or in the process of order allocation, after the order discrete index of the historical order data of a certain online car-hailing vehicle is calculated to be less than or equal to the preset discrete threshold, the model can be updated. Exemplary, under normal circumstances, it may be updated once a month, and after detecting that the order discrete index is less than or equal to the preset discrete threshold, an abnormal update is performed and the update time is recalculated. It should be noted that in order to ensure the reference value of the historical order data, the time range involved in the historical order data can be limited, for example, data within two months before the current time is selected as historical order data for analysis. In this embodiment, no specific limitations are made on the update of the model and the selection of data, and those skilled in the art can reasonably apply them according to actual needs.

[0092] A method for allocating online car-hailing orders provided by an embodiment of the present invention comprehensively considers the driver's driving fatigue level and the round-trip distance involved when the driver does not want to continue to take orders after completing an order and chooses to go home to rest. Specifically, the driver's fatigue level is characterized by calculating the fatigue index, and a prediction model suitable for predicting the order needs of different drivers is selected through the order discrete index to obtain a more accurate prediction result. The return index of the online car-hailing is calculated to quantify the impact of the round-trip distance involved when the driver does not want to continue to take orders after completing an order and chooses to go home to rest. Consideration is taken from multiple aspects to select suitable vehicles for different orders for allocation, effectively improving the efficiency of order allocation and the work quality of drivers and enhancing user experience.

[0093] The embodiment of the present invention further provides an online car-hailing order allocation system, which is used to implement the above-mentioned online car-hailing order allocation method, and specifically includes the following structure:

[0094] An information extraction module is used to determine the order itinerary information according to the order to be allocated after receiving the order to be allocated, including the starting location of the trip, the ending location of the trip and the estimated trip duration, and determine the vehicle screening area based on the order itinerary information;

[0095] The fatigue analysis module is used to obtain the order completion data of multiple online-hailing vehicles in the vehicle screening area that are idle on the same day, and calculate the fatigue index of each online-hailing vehicle based on the order completion data;

[0096] Among them, the fatigue index of each online car-hailing vehicle is calculated based on the order completion data, including:

[0097] The total mileage, total number of orders, average completion time and total rest time corresponding to the order completion data of each online car-hailing vehicle are determined based on the order completion data, and the fatigue index is calculated using the following formula:

[0098] F=w1C1+w2C2+w3C3+w4C4

[0099] Where F is the fatigue index of the online car-hailing service, C1 is the total mileage of passengers dropped off, C2 is the total number of orders, C3 is the average completion time, C4 is the total rest time, and w1, w2, w3 and w4 are the weight parameters of the total mileage of passengers dropped off, the total number of orders, the average completion time and the total rest time, respectively.

[0100] The vehicle analysis module is used to determine whether there is an online-hailing vehicle with a fatigue index less than a preset health index threshold among the multiple online-hailing vehicles in the vehicle screening area that are idle, and when there is no online-hailing vehicle with a fatigue index less than the preset health index threshold among the multiple online-hailing vehicles in the vehicle screening area that are idle, a target vehicle set is constructed based on the multiple online-hailing vehicles in the vehicle screening area that are idle and have fatigue indexes not less than the preset health index threshold, and historical order data and target return locations of each online-hailing vehicle in the target vehicle set are collected;

[0101] The order analysis module is used to calculate the order dispersion index of each online-hailing vehicle in the target vehicle set based on historical order data;

[0102] The order discrete index of each online-hailing vehicle in the target vehicle set is calculated based on historical order data, including:

[0103] The total number of orders, order time span parameters, order frequency and order distribution parameters of each online car-hailing vehicle are determined based on historical order data, and the order discrete index of each online car-hailing vehicle is calculated using the following formula:

[0104] D=u1m1+u2m2+u3m3+u4m4

[0105] Where D is the order dispersion index, m1 is the total number of orders, m2 is the order time span parameter, m3 is the order frequency, m4 is the order distribution parameter, u1, u2, u3 and u4 are the weight parameters of the total number of orders, order time span parameter, order frequency and order distribution parameter respectively;

[0106] Among them, the order time span parameter is the number of days between the first and last orders in the historical order data, and the order distribution parameter is the standard deviation of the orders in the historical order data. The calculation formula is as follows:

[0107]

[0108] Where n is the total number of days corresponding to the historical order data, x i is the number of orders on the i-th day, is the average daily number of orders in the historical order data.

[0109] The order acceptance analysis module is used to select the target prediction model according to the order dispersion index to perform order prediction analysis on each online-hailing vehicle in the target vehicle set and generate the order acceptance confidence of each online-hailing vehicle;

[0110] The order allocation module is used to calculate the return index of each online-hailing vehicle based on the order acceptance confidence and the target return location, and select the online-hailing vehicle with the smallest return index to allocate the orders to be allocated.

[0111] As an optional implementation scheme, the above order allocation module further includes:

[0112] If there is an online-hailing vehicle with a fatigue index lower than the preset health index threshold among multiple idle online-hailing vehicles in the vehicle screening area and the number is 1, the order to be allocated will be allocated to the online-hailing vehicle with a fatigue index lower than the preset health index threshold; if there is and the number is not 1, the online-hailing vehicle closest to the starting location of the trip will be selected for order allocation based on the order itinerary information.

[0113] The above is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. Parts not described in detail in this specification belong to the prior art known to those skilled in the art.

Claims

1. A method for allocating online car-hailing orders, characterized in that: include: After receiving the order to be allocated, determine the order itinerary information according to the order to be allocated, including the starting location of the itinerary, the ending location of the itinerary and the estimated duration of the itinerary, and determine the vehicle screening area based on the order itinerary information; Obtain the order completion data of multiple online-hailing vehicles in the vehicle screening area that are idle on the same day, and calculate the fatigue index of each online-hailing vehicle based on the order completion data; Determine whether there is an online-hailing vehicle whose fatigue index is less than a preset health index threshold among multiple online-hailing vehicles in an idle state in a vehicle screening area; If there is no online-hailing vehicle with a fatigue index less than a preset health index threshold among the multiple online-hailing vehicles in the vehicle screening area that are idle, a target vehicle set is constructed based on multiple online-hailing vehicles in the vehicle screening area that are idle and have fatigue indexes not less than the preset health index threshold, and historical order data and target return locations of each online-hailing vehicle in the target vehicle set are collected; Based on the historical order data, the order dispersion index of each online-hailing vehicle in the target vehicle set is calculated. According to the order dispersion index, the target prediction model is selected to perform a prediction order analysis on each online-hailing vehicle in the target vehicle set, and the order confidence of each online-hailing vehicle is generated. The return index of each online-hailing vehicle is calculated based on the order acceptance confidence and the target return location, and the online-hailing vehicle with the smallest return index is selected to allocate the orders to be allocated; The order dispersion index of each online-hailing vehicle in the target vehicle set is calculated based on historical order data, including: The total number of orders, order time span parameters, order frequency and order distribution parameters of each online car-hailing vehicle are determined based on historical order data, and the order discrete index of each online car-hailing vehicle is calculated using the following formula: D=u1m1+u2m2+u3m3+u4m4 Where D is the order dispersion index, m1 is the total number of orders, m2 is the order time span parameter, m3 is the order frequency, m4 is the order distribution parameter, u1, u2, u3 and u4 are the weight parameters of the total number of orders, order time span parameter, order frequency and order distribution parameter respectively; Among them, the order time span parameter is the number of days between the first and last orders in the historical order data, and the order distribution parameter is the standard deviation of the orders in the historical order data. The calculation formula is as follows: Where n is the total number of days corresponding to the historical order data, x i is the number of orders on the i-th day, is the average daily number of orders in the historical order data.

2. A method for allocating online car-hailing orders according to claim 1, characterized in that: The fatigue index of each online car-hailing vehicle is calculated based on order completion data, including: The total mileage, total number of orders, average completion time and total rest time corresponding to the order completion data of each online car-hailing vehicle are determined based on the order completion data, and the fatigue index is calculated using the following formula: F=w1C1+w2C2+w3C3+w4C4 Where F is the fatigue index of the online car-hailing service, C1 is the total mileage of passengers dropped off, C2 is the total number of orders, C3 is the average completion time, C4 is the total rest time, and w1, w2, w3 and w4 are the weight parameters of the total mileage of passengers dropped off, the total number of orders, the average completion time and the total rest time, respectively.

3. A method for allocating online car-hailing orders according to claim 1, characterized in that: In the process of determining whether there is an online-hailing vehicle whose fatigue index is less than a preset health index threshold among the plurality of online-hailing vehicles in an idle state in the vehicle screening area, the process also includes: If there is an online-hailing vehicle with a fatigue index lower than the preset health index threshold among multiple idle online-hailing vehicles in the vehicle screening area and the number is 1, the order to be allocated will be allocated to the online-hailing vehicle with a fatigue index lower than the preset health index threshold; if there is and the number is not 1, the online-hailing vehicle closest to the starting location of the trip will be selected for order allocation based on the order itinerary information.

4. A method for allocating online car-hailing orders according to claim 1, characterized in that: According to the order dispersion index, the target prediction model is selected to perform order prediction analysis on each online-hailing vehicle in the target vehicle set, and the order confidence of each online-hailing vehicle is generated, including: If the order discrete index of the online car-hailing vehicle is greater than the preset discrete threshold, the personalized prediction model corresponding to the online car-hailing vehicle is selected as the target prediction model, and the target order data of the online car-hailing vehicle within the preset reference range is input into the target prediction model to generate the order confidence of the online car-hailing vehicle; If the order discrete index of the online car-hailing service is not greater than the preset discrete threshold, the general prediction model is selected as the target prediction model, and the target order data of the online car-hailing service within the preset reference range is input into the target prediction model to generate the order confidence of the online car-hailing service.

5. A method for allocating online car-hailing orders according to claim 4, characterized in that: For personalized prediction models and general prediction models, it also includes: Obtain historical order data of multiple online ride-hailing vehicles, extract the second order time series data of each set of historical order data and construct a general training data set, and train a general prediction model based on the general training data set; Based on multiple groups of historical order data, the order discrete index of each online-hailing vehicle is calculated, and multiple groups of historical order data with order discrete index greater than a preset discrete threshold are screened out. The second order receiving time series data of each group of historical order data is extracted and a personalized training data set for each group of historical order data is constructed. Based on the multiple groups of personalized training data sets, personalized prediction models for multiple online-hailing vehicles are trained respectively.

6. A method for allocating online car-hailing orders according to claim 1, characterized in that: The return trip index of each online car-hailing vehicle is calculated based on the order confidence and the target return location, including: The return index is calculated using the following formula: R = (1-P)L, where R is the return index, P is the order confidence, and L is the return mileage, where the return mileage is the driving distance between the end location of the trip and the target return location.

7. A system for allocating online car-hailing orders, characterized in that: The system is used to implement a method for allocating online car-hailing orders as described in any one of claims 1 to 6, comprising: An information extraction module is used to determine the order itinerary information according to the order to be allocated after receiving the order to be allocated, including the starting location of the trip, the ending location of the trip and the estimated trip duration, and determine the vehicle screening area based on the order itinerary information; A fatigue analysis module is used to obtain the order completion data of multiple online-hailing vehicles in the vehicle screening area that are idle on the same day, and calculate the fatigue index of each online-hailing vehicle based on the order completion data; The vehicle analysis module is used to determine whether there is an online-hailing vehicle with a fatigue index less than a preset health index threshold among the multiple online-hailing vehicles in the vehicle screening area that are idle, and when there is no online-hailing vehicle with a fatigue index less than the preset health index threshold among the multiple online-hailing vehicles in the vehicle screening area that are idle, a target vehicle set is constructed based on the multiple online-hailing vehicles in the vehicle screening area that are idle and have fatigue indexes not less than the preset health index threshold, and historical order data and target return locations of each online-hailing vehicle in the target vehicle set are collected; The order analysis module is used to calculate the order dispersion index of each online-hailing vehicle in the target vehicle set based on historical order data; The order acceptance analysis module is used to select the target prediction model according to the order dispersion index to perform order prediction analysis on each online-hailing vehicle in the target vehicle set and generate the order acceptance confidence of each online-hailing vehicle; The order allocation module is used to calculate the return index of each online-hailing vehicle based on the order acceptance confidence and the target return location, and select the online-hailing vehicle with the smallest return index to allocate the orders to be allocated.

8. The online car-hailing order distribution system according to claim 7, characterized in that: For the order allocation module, it also includes: If there is an online-hailing vehicle with a fatigue index lower than the preset health index threshold among multiple idle online-hailing vehicles in the vehicle screening area and the number is 1, the order to be allocated will be allocated to the online-hailing vehicle with a fatigue index lower than the preset health index threshold; if there is and the number is not 1, the online-hailing vehicle closest to the starting location of the trip will be selected for order allocation based on the order itinerary information.

Images