Intelligent management method and system for rental vehicles

Abstract

The present application relates to the technical field of car rental management, in particular to a car rental vehicle intelligent management method and system, comprising the following steps: obtaining the time difference between reservation and use and judging whether to give a warning, counting the number of warnings and adjusting the credit level if the number exceeds the limit, judging the order of renewal and return time to identify abnormal renewal, building behavior nodes and aggregating abnormal statistical behavior chains to give a warning, combining multiple factors to judge the priority of resource requests, and establishing a vehicle planning table. In the present application, short-time frequent reservation behaviors are identified by difference calculation, abnormal renewal is determined, continuous abnormal behaviors are aggregated to deeply model user behavior patterns, credit level records and behavior chain characteristics are combined, user resource request priorities are dynamically adjusted, high-risk behaviors are controlled to interfere with resource scheduling, the warning response capability to abnormal behaviors and the pertinence of resource management are improved, behavior risk constraints and vehicle configuration intelligence are cooperated, and a car rental management system with timeliness, hierarchy and self-adaptive characteristics is built.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of car rental management, in particular to a car rental vehicle intelligent management method and system. BACKGROUND

[0002] The technical field of car rental management includes vehicle information processing, rental business scheduling, user order management, vehicle positioning and state monitoring, etc. The core content is to realize the effective allocation of vehicle resources and the improvement of operation efficiency through the collection of vehicle usage, the reasonable arrangement of rental time and route. Car rental management technology covers real-time information collection, rental process management, platform user interaction control, and rule-based task scheduling, etc. It is a typical resource management information processing technology for business operation.

[0003] Among them, the car rental vehicle intelligent management method refers to a technical scheme for unified planning and management of rental vehicles during the entire rental period, covering vehicle usage state recognition, scheduling logic design based on usage rules, time and location matching control during rental, rental order and vehicle resource matching determination, etc. Usually, the vehicle position information is collected by the position recognition sensor device, combined with the server-side business rule processing, the rental request is matched with resources, and the vehicle instruction is fed back in a remote communication mode. At the same time, the information system is used to jointly judge the rental time window and the idle state of the vehicle and complete the planning and management.

[0004] In the actual management process of existing car rental, the behavior data of user operation is not formed into a unified structured management, and the processing range is limited to the vehicle state and the rental process itself, lacking a continuous abnormal tracking mechanism based on behavior chain construction. In the judgment of multiple non-standard behaviors, the time sequence aggregation judgment standard is not introduced, which makes it difficult to accurately respond when facing short-term frequent operation abnormalities, and it is easy to allow high-frequency abnormal behaviors to escape monitoring, such as the operation of continuously initiating temporary reservations and quickly canceling in a short period of time without forming effective records, which disturbs the resource scheduling plan. In the supervision of the renewal operation, the existing mode only relies on the vehicle state update and the change of the rental time, and lacks the judgment of the actual return intention and the logic before and after the operation, which makes it difficult to identify behaviors that intentionally avoid the return rules. In addition, the user behavior history or behavior risk is not considered in the vehicle allocation, so that high-frequency violators and rule-abiding users are in the same priority sequence in resource allocation, causing high-value resources to be repeatedly occupied, affecting service fairness and scheduling rationality. The behavior path does not form a closed-loop management, and the risk assessment does not build a hierarchical response, resulting in insufficient response capability and precise control capability of the rental management system when facing complex behavior scenarios. SUMMARY

[0005] The purpose of the present application is to solve the shortcomings in the prior art, and to provide a car rental vehicle intelligent management method and system.

[0006] To achieve the above object, the present application adopts the following technical scheme: a car rental vehicle intelligent management method, comprising the following steps:

[0007] S1: obtaining the time difference between the time when the user submits a reservation request on the platform and the starting time of vehicle use, comparing the difference with the minimum reservation advance period, marking the user's reservation warning behavior, and counting the number of reservation warning identification times;

[0008] S2: based on the number of reservation warning identification times, if the identification times are greater than the upper limit of the periodic cumulative warning times, trigger the credit level down adjustment process, bind with the vehicle number, and generate a credit level adjustment record unit;

[0009] S3: obtaining the current state of the unreturned vehicle and recording the user-initiated extension trigger time, determining the return confirmation time, and judging the order of the extension trigger time and the return confirmation time, marking the abnormal extension behavior, and generating an abnormal extension identification tag;

[0010] S4: combining the credit level adjustment record unit and the abnormal extension identification tag, obtaining the corresponding behavior timestamp, vehicle number and behavior type, constructing a behavior node sequence, aggregating abnormal nodes, and counting the number of behavior chain warning nodes;

[0011] S5: based on the number of behavior chain warning nodes, combining the number of reservation warning identification times and the credit level in the credit level adjustment record unit, performing priority judgment on the user's current vehicle resource request, and establishing a car rental vehicle planning table.

[0012] As a further scheme of the present application, the number of reservation warning identification times includes behavior abnormality density, trigger frequency threshold, identification time limit dimension, the credit level adjustment record unit includes level change type, binding behavior node number, associated vehicle unique number, the abnormal extension identification tag includes identification tag type, trigger timestamp, corresponding vehicle state code, the number of behavior chain warning nodes includes continuous node statistics, node behavior type aggregation, node time sequence structure, and the car rental vehicle planning table includes priority sorting result, user request level label, and vehicle allocation basis.

[0013] As a further scheme of the present application, the specific steps of the number of reservation warning identification times are obtained as follows:

[0014] S111: obtaining the time when the user submits a reservation request on the platform and the starting time of vehicle use, calculating the time interval between the reservation request time and the starting time of vehicle use, counting the time difference corresponding to each user reservation data, and generating a vehicle reservation advance time value;

[0015] S112: According to the vehicle reservation advance time value and the shortest reservation advance period reference value, the records with a time difference value less than the shortest reservation advance period reference value are screened, an index of the records is established, and a shortest advance period offset index set is generated;

[0016] S113: According to the shortest advance period offset index set, corresponding record entries are extracted from the original user reservation data, the number of occurrences of the corresponding records in the current time window is counted, the number of advance period abnormal behavior records in the time window is counted, and a reservation warning recognition number is generated.

[0017] As a further scheme of the present application, the specific steps of the credit level adjustment record unit are obtained as follows:

[0018] S211: Based on the reservation warning recognition number and the upper limit of the periodic cumulative warning number, a user index greater than the upper limit of the periodic cumulative warning number is extracted and a set list is formed, and an over-limit warning user index set is generated;

[0019] S212: The over-limit warning user index set is read to locate the credit level state node in the current user database, the existing corresponding credit level of the user is calculated, the credit level of the over-limit warning user is adjusted downward, the current state is confirmed, and a credit level adjustment record is generated;

[0020] S213: According to the credit level adjustment record and the corresponding user index, the current behavior node number is extracted, the vehicle number is extracted from the vehicle allocation record, and the vehicle number is bound to generate a credit level adjustment record unit.

[0021] As a further scheme of the present application, the existing corresponding credit level of the user is calculated, the credit level of the over-limit warning user is adjusted downward, and the formula is used:

[0022]

[0023] The credit level adjustment index value is calculated, and the current state is confirmed, wherein, represents the credit level adjustment value of the user, N ui represents the user U i The number of reservation warning recognition times in a unit period, N max represents the upper limit of the periodic cumulative warning number, L cui represents the user U i The hierarchical position value of the current credit level in the credit level system, T ui represents the user U i The number of days from the registration time of the platform to the present, represents the user U iThe credit level influence factor caused by each default behavior in the last k behavior nodes, N is the total amount of historical behavior nodes included in statistics, and j is the traversal index from 1 to N.

[0024] As a further scheme of the present application, the specific steps for obtaining the abnormal renewal identification label are as follows:

[0025] S311: Obtain all records of the current state of the vehicle that is not returned, filter the renewal request data submitted by the corresponding user, and extract the renewal trigger time corresponding to each renewal request, and match the storage time and return click time of the corresponding vehicle, and generate a set of renewal and return time;

[0026] S312: Based on the vehicle storage time and return click time of each record in the set of renewal and return time, select the maximum value as the return confirmation time, and sequentially judge the return confirmation time and the renewal trigger time, filter out the records with the renewal trigger time earlier than the return confirmation time, and generate renewal time sequence conflict records;

[0027] S313: According to the renewal time sequence conflict record, extract the corresponding record number, construct the behavior state label of the corresponding conflict record in the current task record table, and mark it as abnormal renewal, and generate an abnormal renewal identification label.

[0028] As a further scheme of the present application, the specific steps for obtaining the behavior chain warning node number are as follows:

[0029] S411: Combine the credit level adjustment record unit and the abnormal renewal identification label, extract the corresponding behavior timestamp, vehicle number and behavior type in the two types of records, arrange all records in ascending order according to the behavior timestamp, and aggregate the sorted record information into a single structure to generate a time sequence behavior node sequence;

[0030] S412: Read the behavior type in the time sequence behavior node sequence, mark the abnormal behavior according to the behavior type, count the abnormal marks in adjacent continuous records, filter the abnormal node paragraphs for three times or more, and record the corresponding start and end node numbers, and generate a continuous abnormal segment number group;

[0031] S413: Based on the continuous abnormal segment number group, count the total number of abnormal nodes, and aggregate all abnormal nodes to generate the number of behavior chain warning nodes.

[0032] As a further scheme of the present application, the specific steps for obtaining the rental vehicle planning table are as follows:

[0033] S511: Extract the behavior abnormal record and the reservation abnormal record of the corresponding user based on the number of behavior chain warning nodes and the number of reservation warning identification respectively, combine the user credit level corresponding to each record in the credit level adjustment record unit, and construct a joint abnormal index table indexed by user number;

[0034] S512: Match the joint abnormal index table with the user number in the current vehicle resource request list, calculate the request priority of the user, and add a low priority identifier to the user record with an abnormal identification flag, and mark the remaining records as normal priority, to generate a vehicle request priority label set;

[0035] S513: According to the user request priority and vehicle resource information of each record in the vehicle request priority label set, the request record is connected with the vehicle resource, and the available rental vehicle is deployed and sorted according to the priority size, to generate a rental vehicle planning table.

[0036] As a further scheme of the present application, the user request priority adopts the formula:

[0037]

[0038] for calculation, wherein P k represents the request priority of the user number U k in the joint abnormal index table, X k represents the total number of behavior abnormal nodes of the corresponding user, Y k represents the reservation abnormal times of the corresponding user, represents the adjustment value of the credit level of the corresponding user, G is the total number of all users in the joint abnormal index table, X g represents the number of behavior abnormal nodes of the user in the gth record, Y g represents the reservation abnormal times of the user in the gth record, represents the credit level adjustment value of the user in the gth record.

[0039] The intelligent management system of rental vehicles comprises:

[0040] The reservation warning identification module obtains the reservation request time and the use start time, calculates the time difference, compares with the shortest reservation advance period, judges whether it is a warning behavior, and generates the reservation warning identification times;

[0041] The credit level adjustment module compares the reservation warning identification times with the upper limit of the periodic cumulative warning times to judge whether the credit level is changed, and generates a credit level adjustment record unit;

[0042] The abnormal continuous renting marking module obtains a continuous renting trigger time, a warehousing time and a return click time, judges whether the sequence relationship is abnormal, and generates an abnormal continuous renting identification label if the condition is met;

[0043] The behavior chain aggregation module combines the credit level adjustment recording unit and the abnormal continuous renting identification label, calls a behavior timestamp, a vehicle number and a behavior type to construct a sequence, aggregates continuous abnormal behavior nodes, and generates a behavior chain early warning node quantity.

[0044] The request priority determination module marks a current vehicle request priority based on the behavior chain early warning node quantity, combines the reservation early warning identification times and the credit level adjustment recording unit, and generates a car rental vehicle scheduling table.

[0045] Compared with the prior art, the application has the advantages and positive effects that:

[0046] In the application, short-time frequent reservation behaviors are identified through difference calculation, a behavior early warning mechanism is introduced to enhance identification sensitivity, abnormal continuous renting is determined based on the time sequence relationship between the return confirmation time and the continuous renting operation, the behavior screening capability in the rental process is improved, the node sequence is constructed by using the behavior timestamp, the vehicle number and the behavior type, the continuous abnormal operation is aggregated to generate a behavior chain, the user behavior regularity is deeply modeled, the user resource request priority is dynamically adjusted by combining the credit level record and the behavior chain characteristics, the interference of high-risk behaviors on resource scheduling is controlled, the linkage design of behavior identification precision, time sequence judgment logic and scheduling decision mechanism improves the early warning response capability to abnormal behaviors and the pertinence of resource management, the behavior risk constraint and resource allocation intelligentization are cooperated, a car rental management system with timeliness, hierarchy and self-adaptation characteristics is constructed. BRIEF DESCRIPTION OF DRAWINGS

[0047] Figure 1 The figure is a step flowchart of the application;

[0048] Figure 2 The figure is a system module diagram of the application. DETAILED DESCRIPTION

[0049] In order to make the purpose, technical scheme and advantages of the application clearer, the application is further described in detail below in combination with the drawings and examples. It should be understood that the specific examples described herein are only used to explain the application, and are not used to limit the application.

[0050] In the description of the present application, it should be understood that the terms "length", "width", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation on the present application. In addition, in the description of the present application, the meaning of "a plurality of" is two or more, unless otherwise explicitly and specifically limited.

[0051] Please refer to Figure 1 , the intelligent management method of the rental vehicle, comprising the following steps:

[0052] S1: obtaining the time when the user submits a reservation request on the platform and the start time of vehicle use, calculating the difference between the start time of vehicle use and the reservation request time, comparing the difference with the minimum reservation advance period (such as 2 hours, determined according to industry standards or historical order data analysis), if the difference is less than the minimum reservation advance period, marking the corresponding user reservation behavior as a warning behavior, and counting the number of reservation warning identifications;

[0053] S2: based on the number of reservation warning identifications, if the number of identifications is greater than the upper limit of the periodic cumulative warning number (such as a maximum of 3 warnings per month), trigger the credit level down adjustment process, confirm the user credit level, and bind the current operation behavior node number with the vehicle number, generate a credit level adjustment record unit;

[0054] S3: obtaining the re-rent trigger time initiated by the user when the current state of the vehicle is "not returned", calling the vehicle storage time and the return click time, taking the maximum value of the two as the return confirmation time, judging the order of the re-rent trigger time and the return confirmation time, if the re-rent trigger time is earlier than the return confirmation time, it is marked as "abnormal re-rent", and an abnormal re-rent identification tag is generated;

[0055] S4: combining the credit level adjustment record unit and the abnormal re-rent identification tag, obtaining the corresponding behavior timestamp, vehicle number and behavior type, constructing a behavior node sequence in chronological order, aggregating continuous three times and more abnormal nodes, and counting the number of behavior chain warning nodes;

[0056] S5: based on the number of behavior chain warning nodes, combining the number of reservation warning identifications and the credit level in the credit level adjustment record unit, performing priority judgment on the user's current vehicle resource request, marking the user request with behavior warning or credit adjustment record as low priority, and establishing a rental vehicle planning table.

[0057] The reservation early warning identification number includes behavior abnormality density, trigger frequency threshold, and identification time limit dimension. The credit level adjustment record unit includes level change type, binding behavior node number, and associated vehicle unique number. The abnormality extension identification label includes identification label type, trigger timestamp, and corresponding vehicle state code. The behavior chain early warning node number includes continuous node statistics, node behavior type aggregation, and node time sequence structure. The rental vehicle planning table includes priority sorting result, user request level label, and vehicle allocation basis.

[0058] The specific steps of S1 are as follows:

[0059] S111: Obtain the reservation request time submitted by the user on the platform and the vehicle use start time, calculate the time interval between the reservation request time and the vehicle use start time, count the time difference value corresponding to each user reservation data, and generate a vehicle reservation advance time value;

[0060] To obtain the reservation request time submitted by the user on the platform and the vehicle use start time, the user ID, reservation submission time field, and vehicle reservation start time field corresponding to each reservation record need to be retrieved from the platform database first, and the time value accurate to the minute is extracted through the timestamp format of the time field. In the execution process, the time difference between the reservation request time and the vehicle use start time is calculated. After converting the two time fields into numerical format using a unified time unit (such as minutes), the subtraction operation is performed. For example, in a certain reservation record, the reservation time is April 21, 2024 09:30, and the vehicle use start time is April 22, 2024 08:00. After conversion, they are 202404210930 and 202404220800 respectively. After conversion to minute numbers, they are (9x60+30)+(21x24x60)=3090+30240=33330 minutes and (8x60)+(22x24x60)

[0061] =480+31680=32160 minutes, and the time difference is 1170 minutes. This time difference is the vehicle reservation advance time value. After all records complete the above extraction and calculation process, all vehicle reservation advance time values are summarized in a data list form, and this value is used as the basic data source for subsequent screening and comparison.

[0062] S112: Compare the vehicle reservation advance time value with the shortest reservation advance period benchmark value one by one, filter the records with a time difference value less than the shortest reservation advance period benchmark value, establish a record index, and generate a shortest advance period offset index set;

[0063] According to the comparison between the vehicle reservation advance time value and the minimum reservation advance period reference value, a unified minimum reservation advance period reference value τmin needs to be set first. The reference value refers to the minimum acceptable reservation advance time limit configured in the platform service agreement, such as 720 minutes (i.e. 12 hours). Then, for each reservation record, the vehicle reservation advance time value δi calculated above is extracted and compared with τmin. That is, the judgment operation δi < τmin is performed. For records with a true judgment result, their index value (which can be the database primary key or reservation number) is stored in a list collection to form a minimum advance period offset index set. For example, if a user's reservation advance time value is 500 minutes, which is lower than the 720-minute threshold, it meets the offset condition, and the index value "U000023" is added to the offset index set. Finally, a set such as {U000002, U000023, U000067} is obtained, which will be used as the basis for the next data extraction.

[0064] S113: Extract the corresponding record entries from the original user reservation data according to the minimum advance period offset index set, and count the number of corresponding records in the current time window. Count the number of advance period abnormal behavior records in the time window to generate the reservation warning recognition times;

[0065] According to the minimum advance period offset index set, the corresponding record entries are extracted from the original user reservation data. Each index value in the offset index set is used to locate its complete reservation record entry in the database through a data query statement. The extracted content includes at least user ID, reservation time, use start time, and reservation advance time value fields, and is marked as "abnormal reservation record". Then, according to the current system time T0 and the reservation time of the historical reservation record, the data in the sliding window is counted. The sliding window time length is set to twin (such as 7 days). It is judged whether the reservation time falls within the time interval [T0-twin, T0]. If it falls into, it is counted as an abnormal behavior record. For example, if the current time is April 21, 2024, and the sliding window is 7 days, the records with reservation time between April 14, 2024 and April 21, 2024 are selected and the total number is counted. If there are 5 abnormal records in the offset index set that meet the condition, the generated reservation warning recognition times is 5. This number of times will be used in the subsequent risk modeling and warning output stage.

[0066] The specific steps of S2 are:

[0067] S211: Perform a judgment operation based on the reservation warning recognition times and the upper limit of the periodic cumulative warning times. Extract the user index that is greater than the upper limit of the periodic cumulative warning times and form a collection list to generate the over-limit warning user index set;

[0068] The judgment operation based on the reservation early warning identification times and the upper limit value of the period cumulative early warning times first needs to extract the reservation early warning identification times N of each user in a unit period (such as 30 days) from the output results of the previous stage u , and call the system set period cumulative early warning times upper limit value N max , which can be formulated according to the platform operation specification, and is set to 5 times, and then all users are compared one by one, and the user index U i is called and the corresponding identification times N ui are compared. The comparison operation is executed by judging whether N ui >N max , for each user whose judgment operation result is "true", the index U i is written into a temporary set List e , if users U001, U009, U017 have 7, 6, 3 early warning identification records in the period respectively, the system judges that U001 and U009 satisfy the condition of being greater than the upper limit, and writes List e , U017 is not processed, and after the loop traversal of all user indexes and identification times is completed, the over-limit early warning user index set List e ={U001, U009} is formed.

[0069] S212: Read the over-limit early warning user index set in the current user database to locate the credit level state node, calculate the credit level of the over-limit early warning user according to the existing corresponding credit level of the user, and confirm the current state to generate a credit level down record;

[0070] According to the existing corresponding credit level of the user, the credit level of the over-limit early warning user is calculated, and the formula is as follows:

[0071]

[0072] The credit level down index value is calculated, and the current state is confirmed, wherein, represents the credit level down value of the user, N ui represents the reservation early warning identification times of the user U i in a unit period, N max represents the set period cumulative early warning times upper limit value, represents the current credit level of the user U i in the credit level system (such as A is 1, B is 2, and so on), T ui represents the registration time of the user U i from now on, represents the user U iThe credit level impact factor caused by each default behavior in the past k behavior nodes, N is the total amount of historical behavior nodes included in the statistics, and j is the traversal index from 1 to N;

[0073] The number of pre-warning identification times of the user U001 in a unit period is 7 times, and the upper limit value of the period cumulative warning times is set to 5 times. The upper limit value is derived from the platform operation specification, and based on the statistical results of the normal operation frequency of the user, the proportion of users exceeding 5 times is 9.3%, and the platform sets this as a warning threshold, that is:

[0074] N ui = 7;

[0075] N max = 5, then N ui -N max = 2;

[0076] The current credit level of the user is B level, and in the platform credit level sequence, A level is the highest, corresponding to the value 1, and the B level is 2, so:

[0077] Then

[0078] The user has been registered for 210 days, and the platform calculates the time difference according to the account activation timestamp and the current system time. The timestamp difference is converted by unit conversion to obtain:

[0079] T ui = 210, so

[0080] There are two violation records in the user's last three behavior nodes, and the violation behavior impact factor is generated by the platform behavior risk control model. The model evaluates the severity of the default type, the duration of the behavior, and the intervention frequency, and the quantitative results are 1.2 and 1.8, respectively. The non-default node is marked as 0 points, and the average is:

[0081] N = 3;

[0082]

[0083] Substitute the formula:

[0084]

[0085] The result shows that the credit level downgrading index value of the user U001 is 1.4132, which represents the composite strength level of the credit level adjustment caused by the current behavior and historical state. According to the platform credit level adjustment rule, if the index value is between 1.0 and 2.0, it corresponds to a one-level downgrading, and the credit level will be reduced from B to C. The downgrading information will be recorded in the credit level downgrading record.

[0086] The operation logic of the formula is based on the comprehensive quantification of the behavior violation intensity and the user credit characteristic, where the difference term N ui -N max is used to measure the number of pre-alarm times of the user exceeding the system alert threshold in the current period, reflecting the frequency of abnormal behavior, and the credit level hierarchical value is multiplied, which reflects that the lower the credit level, the more frequent the behavior, and the greater the impact. The product is used as the basic risk factor; the denominator part is subjected to square root operation, which aims to introduce behavior stability reduction weight for users with longer registration time. The longer the registration time, the greater this term, and the overall risk value is lowered, which plays a regulating role; and finally, the average value of the default factor is added, which is used to introduce the historical behavior quality evaluation to avoid adjusting the level only based on the current period behavior, and the overall structure realizes the weight fusion between risk dimensions through weighted multiplication, square root reduction and mean accumulation. The operation result is wrapped by absolute value to ensure its value positive, which has logical consistency.

[0087] S213: According to the credit level down-regulation record and the corresponding user index, the current behavior node number is extracted, the vehicle number is extracted by linking the vehicle allocation record, and the binding is performed to generate the credit level adjustment record unit;

[0088] According to the credit level down-regulation record and the corresponding user index, the current behavior node number is extracted, the vehicle number is extracted by linking the vehicle allocation record, and the binding is performed to generate the credit level adjustment record unit. First, the user ID field U l in each user record in the credit level down-regulation record set Record i is read, and then the latest behavior node number N i of the corresponding user in the behavior log table is located. The behavior node number is generated by the operation track number of the user, such as reservation behavior number, cancellation behavior number, etc. The positioning method is to query the user behavior time sorting field T i , select the item with the maximum time to extract its behavior node number, and then query the vehicle use record related to the user reservation in the vehicle scheduling record table. The corresponding vehicle number C i ​For example, user U001 uses vehicle number C025 in the reservation record, and by querying it is found that vehicle C025 corresponds to behavior node N112, forming the association triple {U001, N112, C025}, which is written into the record unit together with the aforementioned downgraded credit level value, and the complete field item {user ID, behavior node number, vehicle number, original credit level, downgraded credit level, record time} is constructed, for example, generating record unit {U001, N112, C025, B, C, 20240421}. After sequentially completing this operation for all downgraded records, the complete credit level adjustment record unit set is obtained.

[0089] The specific steps of S3 are:

[0090] S311: Obtain all records of vehicles in the current state of not being returned, filter the renewal request data submitted by the corresponding user, and extract the renewal trigger time corresponding to each renewal request, while matching the storage time and return click time of the corresponding vehicle, and generating a renewal and return time set.

[0091] To obtain all records of vehicles in the current state of not being returned, first, the record items with the current state field as "not returned" are filtered from the vehicle state table. The field value is automatically updated by the system vehicle state monitoring module, and the filtering method uses the state field to perform an equal matching operation with the target value "not returned". The vehicle number, associated user ID, and reservation record number are retained in the extraction result. Next, data matching is performed in the renewal request record table based on these vehicle numbers and user IDs. The matching condition is that there is one or more renewal application records under the same vehicle number for the same user ID. After filtering out the successful renewal request data set, the trigger time field T r of each renewal request record is read. Subsequently, the return operation behavior record corresponding to the matched vehicle number is retrieved from the return behavior log table. This involves two key time fields: one is the operation time Tg recorded when the user clicks the return button, and the other is the actual physical storage time Ts of the vehicle. These two items represent the user's subjective initiation of the return operation and the system's record time of confirming that the vehicle has been stored, respectively. Both of them are extracted from the vehicle storage equipment record table and the user operation log table. Finally, the system associates each renewal request record with its matched vehicle number and corresponding T r , Tg, Ts, forming a combination of three time fields, which constitutes the renewal and return time set structure Record t ={T r , Tg, Ts}.

[0092] S312: Based on the vehicle storage time and return click time of each record in the renewal and return time set, select the maximum value as the return confirmation time, and perform a sequential judgment on the return confirmation time and the renewal trigger time to filter out the records where the renewal trigger time is earlier than the return confirmation time, and generate renewal timing conflict records;

[0093] During the execution of selecting the maximum value as the return confirmation time based on the vehicle storage time and return click time of each record in the renewal and return time set, the system sequentially compares Tg and Ts in each group of Record t to determine whether Tg > Ts holds. If it holds, the return confirmation time Tconf = Tg; otherwise, Tconf = Ts. This maximum value selection can be achieved by comparing the magnitudes between two values, and the judgment is completed using simple conditional statements. After execution, each group of renewal records can obtain its corresponding Tconf field. Next, compare this Tconf with the renewal trigger time T r sequentially, that is, perform a conditional judgment on each record to determine whether T r < Tconf holds. If it holds, it means that the renewal occurs before the return is not yet confirmed, there is a timing intersection, and the records that meet this condition will be marked as timing conflict records. The system records the conflict judgment result as "Yes", otherwise as "No". The example is as follows: In the record of a vehicle numbered C101, the renewal trigger time T r is 2024-04-19 14:05, the return click time Tg is 2024-04-19 14:30, and the vehicle storage time Ts is 2024-04-19 14:20. It is judged that the return confirmation time is Tconf = 14:30. Further judgment shows that T r < Tconf holds, and this record is included in the renewal timing conflict set List s . <

[0094] S313: Extract the corresponding record numbers according to the renewal timing conflict records, construct the behavior status label of the corresponding conflict records in the current task record table, and mark them as abnormal renewals to generate abnormal renewal identification labels;

[0095] To extract the corresponding record numbers according to the renewal timing conflict records, the system first needs to retrieve from the renewal timing conflict set List sThe record number or the lease record number Rid bound to each record is extracted, and then the task record total table is entered. The main record position of the conflict record in the task table is quickly located by performing an equal value matching operation on the field Rid. The behavior state label field is attached to this position, and the "abnormal lease renewal" state value is written. At the same time, the content of the existing behavior state field needs to be checked to avoid repeated label stacking. If the field is already "normal", it is directly replaced with "abnormal lease renewal". If other abnormal labels already exist, the labels are merged or updated to a multi-label state, such as "abnormal lease renewal; overdue return". Each marking operation needs to be attached with an operation timestamp Tflag for subsequent operation tracing. Finally, an abnormal lease renewal identification label structure unit Record is formed, which is composed of a record number, a state label, and an operation time. a = {Rid, "abnormal lease renewal", Tflag}. In the example, the record number is R20240419237, the conflict record marking time is 2024-04-21 09:12:00, and the generated label unit is {R20240419237, "abnormal lease renewal", 2024-04-21 09:12:00}. After all record processing is completed, a complete abnormal lease renewal identification label set is formed.

[0096] The specific steps of S4 are as follows:

[0097] S411: Combine the credit level adjustment record unit and the abnormal lease renewal identification label, extract the corresponding behavior timestamps, vehicle numbers, and behavior types in the two types of records, arrange all records in ascending order according to the behavior timestamps, and aggregate the sorted record information into a single structure to generate a time sequence behavior node sequence;

[0098] Combine the credit level adjustment record unit and the abnormal lease renewal identification label. First, extract the corresponding field information from the two record sets. In the credit level adjustment record unit, read the fields containing the user ID, behavior timestamp T1, vehicle number V1, and behavior type identifier L1 (such as "level down"). In the abnormal lease renewal identification label, extract the behavior timestamp T2, vehicle number V2, behavior type identifier L2 (such as "abnormal lease renewal"), and user ID field. After ensuring the uniformity of the fields, perform a field merging operation on the two sets to construct an intermediate behavior event set E = {(T i , V i , L i , U i )}. Then, sort all records in the set E in ascending order according to the timestamp field T i . The sorting method calls T iFields as the sorting primary key, the time format is accurate to the second level comparison, for example, T1 = 2025-04-1909:31:15 and T2 = 2025-04-1909:29:48 are compared and judged, and it is determined that T2 is before T1, after all records are sorted, the index is rearranged according to the sorting result, and a uniform structure body behavior node sequence S = {Node1, Node2, …, Node n} is generated, wherein each Node contains fields {sequence number N i , timestamp T i , vehicle number V i , behavior type L i , user ID U i}, for example, the first record Node1 = {1, 2025-04-1814:10:08, C101, abnormal extension, U001}, the second record Node2 = {2, 2025-04-1817:43:55, C101, grade down, U001}, and finally a complete time sequence behavior node sequence structure S is generated.

[0099] S412: read the behavior type in the time sequence behavior node sequence, mark according to the abnormal behavior identified in the behavior type, count the abnormal marks in adjacent continuous records, filter the abnormal node paragraphs for three times and more, and record the corresponding start and end node numbers, and generate a continuous abnormal segment number group;

[0100] Read the behavior type in the time sequence behavior node sequence, perform content identification operation on the field L i , identify that behaviors such as “abnormal extension” and “violation down” are defined as abnormal behaviors, and the marking method is to add an abnormal flag field B i , if L i is an abnormal behavior type, then B i =1, otherwise 0, the system traverses the behavior node sequence S one by one, performs judgment operation L i ∈ abnormal behavior set E e , if it is satisfied, it is marked as abnormal, after the traversal, the behavior node sequence is converted into an abnormal flag sequence S’ = {Node1(B1), Node2(B2), …}, on this basis, the abnormal flags of adjacent continuous nodes are counted, and B i whether equal to 1 is checked in turn, if three or more B i =1 records are continuously present, then the start node number N s and the end node number N e of the segment are recorded to the number group set List c = {(N s , N eFor example, if B3=B4=B5=B6=1 in four behavior nodes from node 3 to node 6, it is counted as a continuous abnormal behavior segment, and the segment number (3, 6) is written into List_c. The system traverses all node sequences according to this logic, and finally forms a complete continuous abnormal segment number group.

[0101] S413: Based on the continuous abnormal segment number group, the total number of abnormal nodes is counted, and all abnormal nodes are aggregated to generate the number of behavior chain warning nodes.

[0102] Based on the continuous abnormal segment number group, first read all the segment numbers (N s , N e ) from List_c, calculate the number of abnormal nodes in each segment n=N e -N s +1, perform accumulation operation by traversing the number group, and count the total number of abnormal nodes N total =∑(N ei -N si +1). For example, the number group contains (3, 6) and (10, 12), and the total number of abnormal nodes is (6-3+1)+(12-10+1)=4+3=7. Then, filter all nodes with B i =1 in S' to extract all abnormal node information and aggregate to build the abnormal behavior chain set Chain={Node i |B i =1}. Count the number of elements in the set Chain to confirm whether its length is consistent with N total , to ensure statistical consistency. Finally, output the number of abnormal nodes N total as the number of behavior chain warning nodes. For example, the final warning node number is 7.

[0103] The specific steps of S5 are as follows:

[0104] S511: Based on the number of behavior chain warning nodes and the number of reservation warning identification times, extract the behavior abnormal records and reservation abnormal records of the corresponding users, adjust the user credit level corresponding to each record in the credit level adjustment record unit, and construct a joint abnormal index table with user number as index.

[0105] Based on the number of behavior chain warning nodes and the number of reservation warning identification times, extract the behavior abnormal records and reservation abnormal records of the corresponding users. First, extract the fields {user number U i , abnormal node number A i} from the behavior chain warning node set, and take the total number of abnormal nodes corresponding to each user as a quantitative index of behavior abnormality. At the same time, extract the fields {user number U i , reservation abnormality times R i}, representing the number of appointment warning triggers recorded for the user within the cycle window, ensuring that both sets are indexed by the user number for joint matching, and after a successful match, constructing a data structure {U i , A i , R i}, then extracting the fields {user number U i , current credit level L i} from the credit level adjustment record unit, adding the fields to the above structure to form a complete record {U i , A i , R i , L i}, and building a joint abnormality index table T e , in the actual example, if the behavior abnormality node number of U013 is 5 times, the appointment abnormality number is 3 times, and the credit level is C, the corresponding record is {U013, 5, 3, C} written into table T e , which is used for subsequent behavior comprehensive evaluation.

[0106] S512: According to the joint abnormality index table and the user number in the current vehicle resource request list, the request priority of the user is calculated, and the low priority mark is added to the user record with abnormal identification mark, and the rest of the records are marked as normal priority, generating a vehicle request priority label set;

[0107] The request priority of the user is calculated using the formula:

[0108]

[0109] , where P k represents the request priority of the user number U k in the joint abnormality index table, X k represents the total number of behavior abnormality nodes of the corresponding user, Y k represents the number of appointment abnormalities of the corresponding user, represents the adjustment value of the current credit level of the corresponding user (credit levels C, B, A, S are mapped to 3, 2, 1, 0), G is the total number of all users in the joint abnormality index table, X g represents the number of behavior abnormality nodes of the user in the gth record, Y g represents the number of appointment abnormalities of the user in the gth record, represents the credit level adjustment value of the user in the gth record.

[0110] The monitoring source is the data collected by the system behavior chain identification module within the cycle, and the system records that the user's behavior deviates from the expected path 5 times, recorded as X1=5;

[0111] The number of reservation abnormal times is recorded by the reservation early warning identification module. The user triggers the reservation abnormal early warning 3 times in the cycle window, recorded as Y1=3;

[0112] The credit level is automatically adjusted by the credit adjustment unit according to the credit model. The user corresponds to the level C, and the credit level quantization standard is: S→0, A→1, B→2, C→3, so

[0113] The total number of users in the joint abnormal index table is 5, recorded as G=5;

[0114] The number of behavior abnormal nodes and the number of reservation abnormal times of all users in the table are counted respectively, and the field values are as follows: user number U1: X1=5, Y1=3, User number U2: X2=2, Y2=1, User number U3: X3=4, Y3=2, User number U4: X4=1, Y4=0, User number U5:

[0115] X5=0, Y5=1,

[0116] Put the first fraction item into the formula:

[0117]

[0118] Put the sum part into the second fraction item to calculate the numerator:

[0119]

[0120] Calculate the denominator

[0121] Calculate the second fraction value:

[0122]

[0123] Complete the formula:

[0124] P1=|60.5-1.583|=58.917;

[0125] The results show that the request priority characteristic value of user U1 is 58.917, the higher the value, the greater the abnormal intensity, and the lower the priority, so this value is used as a core reference index for low priority identification judgment in the system; wherein the threshold is set to 40, which is based on the distribution statistics of the priority characteristic values of all users in the joint abnormal index table in the period, after 4 consecutive periods of monitoring, the median M and the standard deviation s of the characteristic values of all records are constructed, and the upper limit of the distribution is calculated as M+1.2s, the actual sampling value is calculated as M≈24.5, s≈12.9, and the upper limit threshold is 40.98, which is rounded to 40, the threshold is dynamically adjusted with the fluctuation of the index in different periods, and when a high-density abnormal group appears, the whole is improved, otherwise it falls, the threshold is used to distinguish the priority state of high-risk and low-risk users, if P1>40 triggers the low priority label, generates a record priority field P1=0, and finally corresponds to the annotation result in the vehicle request priority label set generated in step S512.

[0126] The operation logic of the formula reflects the comprehensive weight evaluation of the user abnormal behavior intensity, frequency and credit level, wherein the behavior abnormal node number and the reservation abnormal number have equal importance, therefore, the two are added to form the total abnormal amount, considering the sensitivity of reservation abnormality in influencing resource scheduling, the reservation abnormal number is multiplied by 2 to form the structure of X k +2Y k in the expression, the overall square processing emphasizes the nonlinear amplification effect of abnormal behavior, so that high-frequency abnormality has a significant impact on priority, and the denominator part is processed by taking the square root of the user credit level adjustment value plus 1, the purpose is to nonlinearly compress the credit difference between different levels, avoid the linear dominance of credit level value span on priority, and ensure that high-level users have more reasonable buffer adjustment, the mean value subtracted by the subsequent is the group average state constructed by the aggregation of behavior and credit information of all users, the sum of the total behavior value is constructed by summing the behavior abnormality and the reservation abnormality, and the credit adjustment value plus 1 is constructed, the structure constructs the system overall abnormal baseline in the form of abnormal average / credit adjustment ratio, to balance the relative position of single user abnormal behavior in the group distribution, and finally forms a non-negative value through absolute value operation, to ensure that all priority values have unified directional output in evaluation, which is convenient for subsequent docking with the threshold model to realize identification annotation.

[0127] S513: According to the user request priority in the vehicle request priority label set and the vehicle resource information, the request record is docked with the vehicle resource, and the available rental vehicles are sorted and arranged according to the priority size, and a rental vehicle planning table is generated;

[0128] According to the user request priority of each record in the vehicle request priority label set and the vehicle resource information, the request record is docked with the vehicle resource. First, the available vehicle information set V = {V1, V2, …} is extracted from the vehicle resource pool, and a basic resource number and a state field are assigned to each vehicle. At the same time, the fields {user number U i , request time T r , P i} are read from the vehicle request list. According to the priority identifier P i , all request records are divided into a low-priority group and a normal-priority group. First, the record set with P i = 1 is sorted in ascending order according to the request time T r . The same sorting operation is performed on the records with P i = 0. After sorting the two subsets, the resources are docked in order. The normal group users are preferentially assigned vehicle numbers. The docking method is sequential binding, that is, the first request is bound to the first available vehicle in the order of sorting. If the number of vehicle pools is insufficient, the low-priority group users cannot complete the binding. The system records the request failure state. Finally, the car rental vehicle planning table structure P o = {user number U i , vehicle number V i , priority P i , docking result R o} is generated, where R o = “success” or “failure”. For example, U021 successfully docks the vehicle V102, and the record is {U021, V102, 1, success}. For example, U013 cannot be docked due to low priority and insufficient resources, and the record is {U013, null, 0, failure}.

[0129] Referring to Figure 2 , the car rental vehicle intelligent management system comprises:

[0130] The reservation early warning identification module obtains the reservation request time and the use start time, calculates the time difference, compares it with the shortest reservation advance period, judges whether it is an early warning behavior, and generates a reservation early warning identification number;

[0131] The credit level adjustment module compares the reservation early warning identification number with the upper limit of the periodic cumulative early warning number to determine whether the credit level is changed, and generates a credit level adjustment record unit;

[0132] The abnormal extension marker module obtains the extension trigger time, the storage time and the return click time, judges whether the sequence relationship is abnormal, and generates an abnormal extension identification tag if it is satisfied;

[0133] The behavior chain aggregation module combines the credit level adjustment record unit and the abnormal lease renewal identification tag, calls the behavior timestamp, the vehicle number, and the behavior type to construct a sequence, aggregates continuous abnormal behavior nodes, and generates the behavior chain early warning node quantity;

[0134] The request priority determination module marks the current vehicle request priority based on the behavior chain early warning node quantity, in combination with the reservation early warning identification number and the credit level adjustment record unit, and generates a car rental vehicle scheduling table.

[0135] The above is only a preferred embodiment of the present application, and does not limit the present application in other forms. Any person skilled in the art can use the disclosed technical content to make changes or modifications as equivalent embodiments applied to other fields, but any simple modification, equivalent change, and modification made to the above embodiments without departing from the technical solution content of the present application, in accordance with the technical essence of the present application, still belongs to the protection scope of the technical solution of the present application.

Claims

1. A method for intelligent management of rental vehicles, characterized in that, Includes the following steps: S1: Obtain the time when the user submits the reservation request on the platform and the start time of vehicle use, calculate the difference and compare it with the shortest reservation lead time, mark the user's reservation warning behavior, and count the number of reservation warnings identified. S2: Based on the number of reservation warning recognitions, if the number of recognitions exceeds the upper limit of the cumulative number of warnings in the period, the credit rating downgrade process is triggered, the number of the user's most recent behavior node is located, and the vehicle allocation record is linked to extract the vehicle number and bind it to generate a credit rating adjustment record unit. S3: Obtain the current status of the unreturned vehicle and record the user-initiated renewal trigger time, determine the return confirmation time, judge the order of the renewal trigger time and the return confirmation time, mark abnormal renewal behavior, and generate abnormal renewal identification tags. S4: Combining the credit rating adjustment record unit and the abnormal lease renewal identification tag, obtain the corresponding behavior timestamp, vehicle number and behavior type, construct a behavior node sequence, aggregate abnormal nodes, and count the number of behavior chain warning nodes. S5: Based on the number of warning nodes in the behavior chain, combined with the number of reservation warning recognitions and the credit rating in the credit rating adjustment record unit, priority judgment is performed on the user's current vehicle resource request, and a car rental vehicle planning table is established. The specific steps to obtain the credit rating adjustment record unit are as follows: S211: Based on the number of reservation warnings identified and the upper limit of the cumulative number of warnings in the period, a judgment operation is performed to extract the user indexes that are greater than the upper limit of the cumulative number of warnings in the period and form a set list to generate the over-limit warning user index set; S212: Read the over-limit warning user index set, locate the credit rating status node in the current user database, calculate the downgrade of the credit rating of the over-limit warning user based on the user's existing corresponding credit rating, confirm the current status, and generate a credit rating downgrade record. S213: Based on the credit rating downgrade record and the corresponding user index, locate the most recent behavior node number of the corresponding user in the behavior log table, extract the vehicle number from the vehicle allocation record, bind it, and generate a credit rating adjustment record unit.

2. The intelligent management method for rental vehicles according to claim 1, characterized in that, The reservation warning identification frequency includes abnormal behavior density, trigger frequency threshold, and identification timeliness dimension. The credit rating adjustment record unit includes rating change type, bound behavior node number, and associated vehicle unique number. The abnormal rental renewal identification label includes identification label type, trigger timestamp, and corresponding vehicle status code. The number of behavior chain warning nodes includes continuous node statistics, node behavior type aggregation, and node time series structure. The car rental vehicle planning table includes priority ranking results, user request level label, and vehicle allocation basis.

3. The intelligent management method for rental vehicles according to claim 1, characterized in that, The specific steps to obtain the number of appointment warning recognitions are as follows: S111: Obtain the reservation request time and vehicle usage start time submitted by the user on the platform, calculate the time interval between the reservation request time and the vehicle usage start time, calculate the time difference corresponding to each user reservation data, and generate the vehicle reservation advance time value. S112: Compare each vehicle reservation advance time value with the minimum reservation advance time benchmark value, filter records with a time difference less than the minimum reservation advance time benchmark value, establish a record index, and generate a minimum advance time offset index set. S113: Extract the corresponding record entries from the original user reservation data according to the shortest lead time offset index set, count the number of times the corresponding record appears in the current time window, count the number of abnormal lead time behavior records in the time window, and generate the reservation warning identification count.

4. The intelligent management method for rental vehicles according to claim 1, characterized in that, The calculation of the credit rating downgrade for users who have exceeded the limit warning is based on their existing credit rating, using the following formula: ； Calculate the credit rating downgrade indicator value and confirm the current status, among which, This represents a downgrade in the user's credit rating. This represents the number of appointment alerts identified for user Uᵢ within a set period. This represents the maximum number of warnings that can be issued cumulatively over a set period. This represents the user Uᵢ's current credit rating and its tier position within the credit rating system. This represents the number of days since user Uᵢ registered on the platform. This represents the credit rating impact factor resulting from each default by user Uᵢ in the past k behavioral nodes. This refers to the total number of historical behavior nodes included in the statistics. From 1 to The traversal index.

5. The intelligent management method for rental vehicles according to claim 1, characterized in that, The specific steps to obtain the abnormal lease renewal identification tag are as follows: S311: Obtain all records where the current status of a vehicle is not returned, filter the renewal request data submitted by the corresponding user, extract the renewal trigger time corresponding to each renewal request, and match the entry time and return click time of the corresponding vehicle to generate a set of renewal and return times. S312: Based on the vehicle entry time and return click time of each record in the set of renewal and return times, select the maximum value as the return confirmation time, and judge the order between the return confirmation time and the renewal trigger time to filter out records whose renewal trigger time is earlier than the return confirmation time, and generate renewal time sequence conflict records. S313: Extract the corresponding record number based on the renewal time sequence conflict record, construct the behavior status label of the corresponding conflict record in the current task record table, mark it as abnormal renewal, and generate an abnormal renewal identification label.

6. The intelligent management method for rental vehicles according to claim 1, characterized in that, The specific steps to obtain the car rental vehicle planning schedule are as follows: S511: Based on the number of early warning nodes in the behavior chain and the number of early warning identifications for appointments, extract the corresponding abnormal behavior records and appointment abnormal records for each user. Combine the user credit rating corresponding to each record in the credit rating adjustment record unit to construct a joint abnormality index table indexed by the user ID. S512: Match the user IDs in the joint anomaly index table with the user IDs in the current vehicle resource request list, calculate the user's request priority, add a low priority label to user records with anomaly identification flags, and mark the remaining records as normal priority, thereby generating a vehicle request priority label set. S513: Based on the user request priority and vehicle resource information of each record in the vehicle request priority tag set, the request records are matched with vehicle resources, and the available rental vehicles are allocated and sorted according to their priority to generate a rental vehicle planning table.

7. The intelligent management method for rental vehicles according to claim 6, characterized in that, The priority of a user's request is determined by the formula: ； Calculations are performed, in which, User ID represented in the Joint Anomaly Index Table is Request priority This represents the total number of nodes with abnormal behavior for the corresponding user. This represents the number of times the corresponding user's appointments have been abnormal. This represents the adjustment value for the corresponding user's current credit rating. This represents the total number of all users in the combined anomaly index table. Representing the The number of abnormal user behavior nodes in the record. Representing the The number of times a user's appointment was abnormal in the record. Representing the The credit rating adjustment value for the user in the record.

8. A car rental vehicle intelligent management system, characterized in that, The system is used to implement the intelligent vehicle management method for rental cars according to any one of claims 1-7, the system comprising: The reservation warning identification module obtains the reservation request time and the usage start time, calculates the time difference, compares it with the shortest reservation lead time, determines whether it is a warning behavior, and generates the reservation warning identification count. The credit rating adjustment module compares the number of appointment warnings identified with the upper limit of the cumulative number of warnings over a period to determine whether a credit rating change is triggered and generates a credit rating adjustment record unit. The abnormal renewal tagging module obtains the renewal trigger time, the entry time and the return click time, and determines whether the sequence relationship is abnormal. If it is, an abnormal renewal identification tag is generated. The behavior chain aggregation module combines the credit rating adjustment record unit with the abnormal lease renewal identification tag, calls behavior timestamp, vehicle number, and behavior type to construct a sequence, aggregates continuous abnormal behavior nodes, and generates the number of behavior chain warning nodes. The request priority determination module, based on the number of early warning nodes in the behavior chain, combined with the number of reservation early warning recognitions and the credit rating adjustment record unit, marks the current vehicle request priority and generates a car rental vehicle dispatch table.

Images