A user charging pile recommendation method and device based on collaborative filtering

Abstract

The present invention relates to a collaborative filtering-based method and device for recommending charging piles for users. The method is characterized by grouping user interactions with charging piles based on spatiotemporal factors, and using a collaborative neural network to calculate a recommended list of charging piles for each group. The method comprises the following steps: obtaining user-charging pile interaction information and their geographic locations based on interaction information between the user and the charging pile, and generating a binary matrix of user-charging pile and charging pile-user relationships; importing the binary matrix as an input vector into the collaborative neural network to calculate a recommended list of charging piles for the user; and matching the recommended list of charging piles for the user with the user's current SOC alarm location to recommend charging piles that meet the current SOC alarm location conditions. This allows charging pile recommendations to better meet the user's personalized selection needs.

Description

Technical Field

[0001] The present invention relates to the field of intelligent recommendation technology, and in particular to a method and device for recommending user charging piles based on collaborative filtering. Background Art

[0002] In recent years, more and more electric vehicles have entered the market, and the number of public charging piles has also increased day by day. Due to the charging fees and charging speed issues of charging piles, users will not be able to quickly select a charging pile when the SOC alarm sounds.

[0003] Related technologies obtain prediction parameters for charging stations within a preset distance range. Based on the availability probability of each charging station within the predicted parameter range, they recommend charging stations that meet the load demand. However, these technologies can only predict the availability of certain charging stations, which influences user choices and fails to meet users' personalized selection needs. For example, when selecting a charging station, users may prefer to save money and be relatively indifferent to time costs, or they may prefer to save time and be relatively indifferent to cost. Summary of the Invention

[0004] The embodiments of the present invention provide a method and device for recommending charging piles to users based on collaborative filtering, which can make the recommendation of charging piles more in line with the personalized selection needs of users.

[0005] On the one hand, an embodiment of the present invention provides a user charging pile recommendation method based on collaborative filtering, characterized in that the interaction behaviors between users and charging piles are grouped according to temporal and spatial factors, and a collaborative neural network is used to calculate a charging pile recommendation list based on different groups. The method includes the following steps:

[0006] Based on the interaction information between the user and the charging pile, the interaction status and interaction geographic location of the user and the charging pile are obtained, and a binary matrix of the user to the charging pile and the charging pile to the user is generated;

[0007] Importing the binarized matrix as an input vector into the collaborative neural network to calculate a charging pile recommendation list for the user;

[0008] After matching the recommended list of charging piles for the user with the user's current SOC alarm location, a charging pile that meets the current SOS alarm location condition is recommended.

[0009] In some embodiments, the interaction status includes whether interaction is generated or not;

[0010] The method of generating a binary matrix of user-to-charging pile and charging pile-to-user includes the following steps:

[0011] In the same group, according to the interactive geographic location, the interaction between the user and the charging pile and the charging pile and the user are represented by one-hot encoded binary vectors, and recorded as user vector and charging pile vector respectively, and

[0012] The user vector includes the user's interaction with all charging piles at the interactive geographic location.

[0013] The charging pile vector includes the interaction status of the charging pile with all users.

[0014] In some embodiments, the binarized matrix is ​​imported as an input vector into the collaborative neural network to calculate a charging pile recommendation list for the user, comprising the steps of:

[0015] Importing the user vector and the charging pile vector together as input layer vectors into the embedding layer of the collaborative neural network for matrix decomposition to obtain a dense vector;

[0016] The NCF prediction model based on the collaborative neural network maps the dense vector into a prediction score, and the NCF prediction model is expressed as:

[0017] in, is the prediction score, P∈R M×K And represents the user's potential factor matrix, Q∈R N×K And represents the potential factor matrix of the charging pile, θ f are the parameters of the interaction function f, f is the collaborative neural network and f is expressed as:

[0018] in, Represents the input user vector The potential feature vector of user u obtained after the Embedding layer is: Represents the input charging pile vector The potential feature vector of charging pile i obtained after the Embedding layer is: represents the mapping function of the xth layer in the neural collaborative filtering layer, Represents the mapping function of the output layer;

[0019] A list of charging pile recommendations for the user is determined based on the predicted scores.

[0020] In some embodiments, before determining the charging pile recommendation list for the user based on the predicted scores, the method further includes the following steps:

[0021] A prediction target value is set and the NCF prediction model is trained by minimizing the mean square error between the prediction target value and the prediction score.

[0022] In some embodiments, the training of the NCF prediction model by minimizing the mean square error between the predicted target value and the predicted score comprises the steps of:

[0023] The objective function is set according to a first formula, wherein the first formula includes:

[0024]

[0025]

[0026] Among them, y ui To predict the target value, is the prediction score, L is the objective function, y represents the observation data in the interaction matrix generated by the interaction behavior between the user and the charging pile, and y - represents a negative sample of the observed data;

[0027] Solve the objective function to obtain the optimal parameter values ​​of the NCF prediction model.

[0028] In some embodiments, the user's most recent one or more interaction behaviors are used as the predicted target value.

[0029] In some embodiments, grouping the interaction behaviors between users and charging piles according to time and space factors includes the steps of:

[0030] Grouping by region is used to group the interaction behaviors between users and charging piles in the same region into the same large group based on regional divisions;

[0031] After the grouping by area, grouping by different time intervals may be optionally performed, and the grouping by different time intervals includes grouping the interaction behaviors between users and charging piles within the same time interval into the same group;

[0032] The grouping of large groups or small groups is taken as the final grouping result.

[0033] In some embodiments, matching the charging pile recommendation list for the user with the user's current SOC alarm location and recommending a charging pile that meets the current SOS alarm location condition includes the steps of:

[0034] Get the user's current SOC alarm location;

[0035] Calculating the distance between the user and each recommended charging pile based on the charging pile recommendation list;

[0036] According to the reachable distance of the user's remaining power, the charging pile that meets the current SOS alarm location is selected from the recommended charging piles and pushed.

[0037] In some embodiments, matching the charging pile recommendation list for the user with the user's current SOC alarm location and recommending a charging pile that meets the current SOS alarm location condition includes the steps of:

[0038] If the current SOC alarm location is an area that the user has never been to, the user's current location information is collected, and the corresponding area group is searched according to the user's current location information, and recommendations are made according to the charging pile recommendation list under the corresponding area group.

[0039] On the other hand, an embodiment of the present invention also provides a device, characterized in that the device includes: at least one processor; and a memory coupled to the at least one processor, the memory containing instructions stored therein, and the instructions are loaded and executed by the processor to implement any one of the methods described in the method embodiments.

[0040] The embodiment of the present invention can group the interaction behaviors between users and charging piles according to time and space factors, and use collaborative neural networks to calculate the recommended list of charging piles according to different groups. It can group according to factors such as geographical location or time conditions (such as season, time period) and calculate the recommended list separately, which can reduce the CPU load and enhance the accuracy of the algorithm. At the same time, many important factors that affect user choices, such as geographical location (including city, region), are incorporated into the neural network algorithm as input weight factors, and a list of charging piles that the user likes can be pushed when the user has an SOC alarm at an unfamiliar location. After calculating the recommended charging pile list, the list will be matched with the user's SOC alarm location, and the charging piles that the user cannot reach will be eliminated before being pushed. This can make the recommendation of charging piles more in line with the user's selection logic and the results more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0042] Figure 1 A flowchart of a method for recommending charging piles to users based on collaborative filtering according to an embodiment of the present invention is provided;

[0043] Figure 2 A schematic diagram of a collaborative neural network provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0044] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, 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. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0045] like Figure 1 As shown, an embodiment of the present invention provides a method for recommending charging piles to users based on collaborative filtering, including the following steps:

[0046] S100: Based on the interaction information between the user and the charging pile, the interaction status and interaction geographic location of the user and the charging pile are obtained, and a binary matrix of the user to the charging pile and the charging pile to the user is generated;

[0047] S200: Importing the binarized matrix as an input vector into the collaborative neural network to calculate a charging pile recommendation list for the user;

[0048] S300: After matching the recommended list of charging piles for the user with the user's current SOC alarm location, a charging pile that meets the current SOS alarm location conditions is recommended.

[0049] It should be noted that prior to S100, user interactions with charging stations are grouped based on spatiotemporal factors. This recommendation method uses a collaborative neural network to calculate a list of recommended charging stations for each group. Spatiotemporal factors can be understood as time and / or space factors. Time factors can include seasonal factors or time period factors, and spatial factors can include geographic location factors, such as using different cities as geographic location factors for grouping.

[0050] In S100, the interaction information between users and charging piles comes from the electric vehicle charging history data uploaded by the TSP (Telematics Service Provider). This historical data is filtered and combined to obtain the basic data for charging analysis. It is believed that the basic data has undergone data cleaning processes such as parsing, standardization, and outlier processing to screen out miscellaneous and useless information.

[0051] It is understandable that the collaborative neural network in S200 can be as follows Figure 2 The multi-layer neural network shown is used to calculate user preferences for charging stations based on the similarities between users and use this as a recommendation result (charging station recommendation list). This reduces uncertainty by calculating user preferences and avoids relying solely on charging station information to push recommendations, which can overlook user subjective preferences.

[0052] It should be noted that the SOC in S300 refers to the state of charge of the electric vehicle, which indicates the ability of the battery to continue working.

[0053] In an embodiment of the present invention, the interaction behaviors between users and charging piles are grouped according to spatiotemporal factors, and a collaborative neural network is used to calculate a recommended list of charging piles according to different groups. Grouping can be performed according to factors such as geographical location or time conditions (such as season, time period), and the recommended list can be calculated separately, which can reduce the CPU load and enhance the accuracy of the algorithm. At the same time, many important factors that affect user choices, such as geographical location (including city, region), are incorporated into the neural network algorithm as input weight factors, and a list of charging piles that the user likes can be pushed when the user has an SOC alarm at an unfamiliar location. After the recommended charging pile list is calculated, the list will be matched with the user's SOC alarm location, and the charging piles that the user cannot reach will be eliminated before being pushed. This can make the recommendation of charging piles more in line with the user's selection logic, and the results more accurate.

[0054] Furthermore, the interaction behaviors between users and charging piles are grouped according to time and space factors, including the following steps:

[0055] S001: Grouping by region, which is used to group the interaction behaviors of users and charging piles in the same region into the same large group according to the regional division;

[0056] S002: After the grouping by area, optionally grouping by different time intervals is performed, and the grouping by different time intervals includes grouping the interaction behaviors between users and charging piles within the same time interval into the same group;

[0057] S003: The grouping of large groups or small groups is taken as the final grouping result.

[0058] It is understandable that the grouping methods of S001 and S002 can be either one or the other, or S001 can be grouped in a progressive manner, and then S002 can be grouped into small groups under the condition of the large grouping, and finally the small grouping can be used as the grouping result.

[0059] In a specific embodiment, three groupings are performed. The first grouping is performed according to different cities (i.e., regions). The second grouping is performed based on this grouping result, and the grouping is performed according to the charging period (for example, charging during the day or charging at night). On this basis, the third grouping is performed, that is, grouping is performed according to the season (such as summer or winter), and the final grouping is used as the grouping result for subsequent calculations.

[0060] This embodiment of the present invention addresses the issue of excessive CPU load when calculating the relationship between each user and each charging station. Due to the large number of users and charging stations (for a case with 1 million vehicles and 100,000 charging stations), a 1 million × 100,000 matrix is ​​required for subsequent calculations. By proposing grouping and rationalizing the grouping based on various factors that influence user preference, this effectively reduces CPU load while improving calculation accuracy.

[0061] In some embodiments, the interaction status includes whether interaction occurs and whether interaction does not occur. In S100, a binary matrix of user-to-charging pile and charging pile-to-user is generated, including the steps of:

[0062] In the same group, according to the interactive geographic location, the user's interaction with the charging pile and the charging pile's interaction with the user are represented by one-hot encoded binary vectors, and recorded as user vectors and charging pile vectors respectively. The user vector includes the user's interaction with all charging piles in the interactive geographic location, and the charging pile vector includes the interaction of the charging pile with all users.

[0063] In a specific embodiment, the interaction between the user and the charging pile is recorded as 1, and the non-interaction is recorded as 0. It can be concluded that the interaction of user X1 with the charging pile (Y1, Y2, Y3, ..., Yn) at the geographical location is represented as (1, 0, 0, ..., 1); similarly, the interaction between charging pile Y1 and user (X1, X2, X3, ..., Xn) is represented as (0, 1, 0, ..., 1). Since it is impossible for a user to interact with all charging piles, there are many values ​​0 in the above vector, that is, it is a sparse vector.

[0064] In some embodiments, in S200, the binarized matrix is ​​imported as an input vector into the collaborative neural network to calculate a charging pile recommendation list for the user, including the steps of:

[0065] S210: Importing the user vector and the charging pile vector as input layer vectors into the embedding layer of the collaborative neural network to perform matrix decomposition to obtain a dense vector;

[0066] S220: Mapping the dense vector into a prediction score based on the NCF prediction model of the collaborative neural network, and the NCF prediction model is expressed as:

[0067] in, is the prediction score, P∈R M×K And represents the user's potential factor matrix, Q∈R N×K And represents the potential factor matrix of the charging pile, θ f are the parameters of the interaction function f, f is the collaborative neural network and f is expressed as:

[0068] in, Represents the input user vector The potential feature vector of user u obtained after the Embedding layer is: Represents the input charging pile vector The potential feature vector of charging pile i obtained after the Embedding layer is: represents the mapping function of the xth layer in the neural collaborative filtering layer, Represents the mapping function of the output layer;

[0069] S230: Determine a recommended list of charging stations for the user based on the predicted scores.

[0070] It should be noted that if Figure 2 The multi-layer neural network (i.e. collaborative neural network) shown in the figure has an input layer at the bottom, which contains two feature vectors and They are used to describe user u and charging pile i respectively; the input consists of only a user vector and a charging pile vector, which are binary sparse vectors encoded with one-hot. Above the input layer is the Embedding layer, which is a fully connected layer used to map the binary sparse vector of the input layer into a dense vector. The dense vector of the user (charging pile) obtained by it can be regarded as a potential feature vector used to describe the user (charging pile). The output vector (dense vector) of the embedding layer of the user and the charging pile is then sent to the NCF prediction model of the multi-layer neural network, which includes a neural collaborative filtering layer (Neural CF Layer), which is used to map the potential feature vector into a prediction score (Score). Each layer in the neural collaborative filtering layer can be considered to be used to explore certain implicit structures in the user-charging pile interaction. The dimension of the output layer (Layer X) of the neural collaborative filtering layer determines the capability of the model. The final output is the prediction score

[0071] Preferably, the charging interaction data of users in the past six months is screened, the geographical locations visited by users in the past six months are counted, the number of interactions with charging piles at each location is calculated, the weights of the locations are sorted according to the number of times, and the location weights are brought into the calculation in the collaborative neural network to achieve the purpose of reducing the CPU load and enhancing the accuracy of the algorithm.

[0072] In some embodiments, before S300, the process further includes:

[0073] S240: Setting a prediction target value and training the NCF prediction model by minimizing the mean square error between the prediction target value and the prediction score.

[0074] Furthermore, S240 includes the steps of:

[0075] S241: Setting an objective function according to a first formula, wherein the first formula includes:

[0076]

[0077] Among them, y ui To predict the target value, is the prediction score, L is the objective function, y represents the observation data in the interaction matrix generated by the interaction behavior between the user and the charging pile, and y - represents a negative sample of the observed data;

[0078] S242: Solve the objective function to obtain the optimal parameter values ​​of the NCF prediction model.

[0079] Preferably, the user's most recent interactive behavior is used as the predicted target value.

[0080] It can be understood that according to the user's most recent one or more selections of the charging pile, the label is used as the target value predicted by the algorithm. For the predicted results, it is first determined whether the label appears in the result list, and then it is determined whether the order of appearance of the label in the list is high, so as to achieve the purpose of accurate prediction.

[0081] The target values ​​and other weights involved in the embodiments of the present invention can be iteratively modified based on the continuously accumulated electric vehicle charging data.

[0082] In some embodiments, S300 includes the steps of:

[0083] S310: Obtain the user's current SOC alarm position;

[0084] S320: Calculating the distance between the user and each recommended charging pile based on the charging pile recommendation list;

[0085] S330: Filter out the charging piles that meet the current SOS alarm location from the recommended charging piles according to the distance range where the user's remaining power can reach, and push the charging piles.

[0086] In this embodiment of the present invention, the charging pile positions in the matching list are matched with the user's SOC warning position, the distance between the two is calculated, and the charging piles with unreachable remaining SOC are excluded. The charging piles in the remaining list are the recommended charging piles for the user.

[0087] In some embodiments, when S300 pushes the message to the user, if the current SOC alarm location is an area that the user has never been to, the user's current location information is collected, and the corresponding area group is searched based on the user's current location information, and recommendations are made according to the charging pile recommendation list under the corresponding area group.

[0088] It is understandable that even though the user's current location is an area they have never visited, the preferred charging station list in this embodiment of the present invention has already been calculated by incorporating geographic location as a weighting factor into the collaborative neural network algorithm. Therefore, the preferences of other users in any unfamiliar area can be calculated. Therefore, the calculation results (charging station recommendation list) for that area group can be directly applied to this user.

[0089] In some embodiments, when the next SOC alarm of the vehicle is predicted, a reminder of "charging piles of interest" will be actively pushed to the driver during driving. When the vehicle driver selects the "charging piles of interest" reminder list on the vehicle side, the predicted charging pile list will be pushed to the user, providing the user with charging options.

[0090] The beneficial effects brought about by the technical solution of the present invention include:

[0091] The present invention predicts the user's preference for charging piles with which he has not interacted based on the interaction data between the user and the charging pile in combination with a collaborative neural network, which better matches the user's subjective choice. In addition, the charging pile recommendation provided by this method can push the user's favorite charging pile when the user receives an SOC alarm at an unfamiliar location. At the same time, this method can repeatedly optimize the model recommendation results by continuously inputting data.

[0092] Compared to related technologies that perform calculations for multiple users on a platform (such as a supermarket), without geographical or temporal distinctions, and only consider the correlation between users and products, the embodiments of the present invention not only consider the interaction between users and charging piles, but also consider the geographical location (such as the city) as an important factor affecting user choices and incorporate it into subsequent calculations. After calculating the list of recommended charging piles, the list will be matched with the user's SOC alarm location, and the charging piles that the user cannot reach will be eliminated before being pushed. This can make the recommendation of charging piles more in line with the user's selection logic and the results more accurate.

[0093] On the other hand, an embodiment of the present invention also provides a device comprising: at least one processor; and a memory coupled to the at least one processor, the memory containing instructions stored therein, the instructions being loaded and executed by the processor to implement any one of the methods described in the method embodiments.

[0094] It can be understood that the device can achieve the technical effects that can be achieved by any method in the aforementioned method embodiments, and the scope of its technical effects is equivalent.

[0095] In the description of the present invention, it should be noted that the orientation or positional relationship indicated by the terms "upper" and "lower" is based on the orientation or positional relationship shown in the accompanying drawings, and is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention. Unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be internal communication between two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to the specific circumstances.

[0096] It should be noted that, in the present invention, relational terms such as "first" and "second" are merely used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply the existence of any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprising a ..." does not exclude the presence of other identical elements in the process, method, article or device comprising the element.

[0097] The foregoing description is intended only to provide specific embodiments of the present invention, which will enable those skilled in the art to understand and implement the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not intended to be limited to the embodiments shown herein, but is intended to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A user charging pile recommendation method based on collaborative filtering, characterized in that: The interaction behaviors between users and charging piles are grouped according to time and space factors, and a collaborative neural network is used to calculate a charging pile recommendation list according to different groups. The method includes the following steps: Based on the interaction information between the user and the charging pile, the interaction status and interaction geographic location of the user and the charging pile are obtained, and a binary matrix of the user to the charging pile and the charging pile to the user is generated; Importing the binarized matrix as an input vector into the collaborative neural network to calculate a charging pile recommendation list for the user; After matching the charging pile recommendation list for the user with the user's current SOC warning location, recommending a charging pile that meets the current SOC warning location condition; The interaction situation includes whether interaction occurs or not; The method of generating a binary matrix of user-to-charging pile and charging pile-to-user includes the following steps: In the same group, according to the interactive geographic location, the interaction between the user and the charging pile and the charging pile and the user are represented by one-hot encoded binary vectors, and recorded as user vector and charging pile vector respectively, and The user vector includes the user's interaction with all charging piles at the interactive geographic location. The charging pile vector includes the interaction status of the charging pile with all users; The method of grouping the interaction behaviors between users and charging piles according to time and space factors includes the following steps: Grouping by region is used to group the interaction behaviors between users and charging piles in the same region into the same large group based on regional divisions; After the grouping by area, grouping by different time intervals is performed, and the grouping by different time intervals includes grouping the interaction behaviors between users and charging piles in the same time interval into the same group; The grouping of large groups or small groups is regarded as the final grouping result; After matching the charging pile recommendation list for the user with the user's current SOC warning position, a charging pile that meets the current SOC warning position condition is recommended, including the steps of: If the current SOC alarm location is an area the user has never been to, the user's current location information is collected, and the corresponding area group is searched based on the user's current location information, and a charging pile recommendation list under the corresponding area group is recommended; A weighting factor using geographic location as input is incorporated into the neural network algorithm, wherein the geographic location includes a city or a region.

2. The method for recommending charging piles for users based on collaborative filtering according to claim 1, characterized in that: The binarized matrix is ​​imported as an input vector into the collaborative neural network to calculate a charging pile recommendation list for the user, including the steps of: Importing the user vector and the charging pile vector together as input layer vectors into the embedding layer of the collaborative neural network for matrix decomposition to obtain a dense vector; The NCF prediction model based on the collaborative neural network maps the dense vector into a prediction score, and the NCF prediction model is expressed as: ,in, is the prediction score, represents the user’s latent factor matrix, represents the potential factor matrix of the charging pile, are the parameters of the interaction function f, For the collaborative neural network Expressed as: ,in, Represents the input user vector The potential feature vector of user u obtained after the Embedding layer is: Represents the input charging pile vector The potential feature vector of charging pile i obtained after the Embedding layer is: represents the mapping function of the xth layer in the neural collaborative filtering layer, Represents the mapping function of the output layer; A charging pile recommendation list for the user is determined based on the predicted scores.

3. The method for recommending charging piles based on collaborative filtering according to claim 2, characterized in that: Before determining the charging pile recommendation list for the user according to the predicted scores, the method further includes the following steps: A prediction target value is set and the NCF prediction model is trained by minimizing the mean square error between the prediction target value and the prediction score.

4. The method for recommending charging piles for users based on collaborative filtering according to claim 3, characterized in that: The training of the NCF prediction model by minimizing the mean square error between the predicted target value and the predicted score comprises the steps of: The objective function is set according to a first formula, wherein the first formula includes: ， ， in, To predict the target value, is the prediction score, is the objective function, represents the observation data in the interaction matrix generated by the interaction behavior between the user and the charging pile, and represents a negative sample of the observed data; Solve the objective function to obtain the optimal parameter values ​​of the NCF prediction model.

5. The method for recommending charging piles for users based on collaborative filtering according to claim 3, characterized in that: The user's most recent one or more interaction behaviors are used as the predicted target value.

6. The method for recommending charging piles for users based on collaborative filtering according to claim 1, characterized in that: After matching the charging pile recommendation list for the user with the user's current SOC warning position, a charging pile that meets the current SOC warning position condition is recommended, including the steps of: Get the user's current SOC alarm location; Calculating the distance between the user and each recommended charging pile based on the charging pile recommendation list; According to the distance range where the user's remaining power can reach, the charging pile that meets the current SOC warning position is selected from the recommended charging piles and pushed.

7. A device, characterized in that The device includes: at least one processor; and a memory coupled to the at least one processor, wherein the memory contains instructions stored therein, and the instructions are loaded and executed by the processor to implement the method according to any one of claims 1 to 6.

Images