Product recommendation methods, devices, equipment, storage media, and program products.

Abstract

This application provides a product recommendation method, apparatus, device, storage medium, and program product. This application relates to the field of recommendation system technology. The method includes: acquiring target features of multiple candidate fund products, the target features including indicator features used to evaluate candidate fund products in historical periods and personalized features of recommenders who recommended candidate fund products in historical periods; for each candidate fund product, predicting the risk-return outcome of each candidate fund product based on the target features and using a product evaluation model; and determining a target fund product for recommendation from the multiple candidate fund products based on the predicted risk-return outcome. The method provided by this application overcomes the problem that existing technologies cannot accurately evaluate fund products, thus failing to accurately and effectively recommend fund products to users.

Description

Technical Field

[0001] This application relates to the field of recommendation system technology, and in particular to a product recommendation method, apparatus, device, storage medium, and program product. Background Technology

[0002] In wealth management, product management and product selection are among the key areas for improvement. In recent years, as people have placed increasing emphasis on wealth management, the market has seen a wide variety of products (such as mutual funds). To better provide clients with quality products, product evaluation is constantly being optimized and is developing rapidly.

[0003] Currently, methods used to evaluate products include the Sharpe ratio, Treynor ratio, and Jensen alpha (also known as the Jensen index). However, the Sharpe ratio, based on the Capital Asset Pricing Model (CAPM) assumptions and using the risk-free rate as a benchmark, adjusts for total risk, thus leading to some bias in evaluating asset returns. The Treynor ratio, where a higher ratio indicates better asset performance, fails to consider the existence of unsystematic risk in actual fund portfolios, resulting in an overly idealistic measurement of risk. Jensen alpha, where a higher ratio indicates greater excess returns and better asset performance, and vice versa, does not necessarily reflect excess returns for either the overall market or individual assets.

[0004] Therefore, existing technologies cannot accurately evaluate fund products, and thus cannot accurately and effectively recommend fund products to users. Summary of the Invention

[0005] This application provides a product recommendation method, apparatus, device, storage medium, and program product to overcome the problem that the prior art cannot accurately evaluate fund products, and therefore cannot accurately and effectively recommend fund products to users.

[0006] In a first aspect, embodiments of this application provide a product recommendation method, the method comprising:

[0007] Acquire target features of multiple candidate fund products, including indicator features used to evaluate candidate fund products in historical periods and personalized features of recommenders who recommended candidate fund products in historical periods.

[0008] For each of the candidate fund products, based on the target characteristics, the return and risk outcomes of each candidate fund product are predicted using a product evaluation model;

[0009] Based on the predicted return and risk results, a target fund product for recommendation is determined from the plurality of candidate fund products.

[0010] In one possible design, obtaining the target features of multiple candidate fund products includes:

[0011] Acquire feature-related data for multiple candidate fund products. The feature-related data includes indicator data used to evaluate candidate fund products in historical periods and personalized data of recommenders who recommended candidate fund products in historical periods. The indicator data includes at least one of the following: return rate indicator, Sharpe ratio, Treynor ratio, excess return, risk indicator, and investment type of fund product. The personalized data of recommenders includes at least one of the following: the recommender's basic personal data and data reflecting their capabilities.

[0012] For each candidate fund’s feature-related data, extract the indicator features corresponding to the indicator data and the personalized features of the recommender corresponding to the recommender’s personalized data.

[0013] The target features consist of indicator features and the personalized features of the recommender.

[0014] In one possible design, the step of extracting the indicator features corresponding to the indicator data and the personalized features of the recommender corresponding to the recommender's personalized data includes:

[0015] Principal component regression analysis was performed on the aforementioned return rate indicators, Sharpe ratio, Treynor ratio, excess return, and risk indicators to obtain the regression analysis results.

[0016] The investment type and the recommender's basic personal data are uniquely encoded to obtain encoded data;

[0017] The capability characteristics are obtained by processing the capability-reflecting data using the least squares method.

[0018] The principal component regression analysis is used to reduce the dimensionality of the regression analysis results, coded data, and capability features to obtain the target features.

[0019] In one possible design, predicting the risk-return outcomes of each candidate fund product based on the target characteristics using a product evaluation model includes:

[0020] The target features are input into the product evaluation model. By continuously predicting the returns of each fund in the candidate fund products and weighting the returns of each fund, the predicted returns of the candidate fund products in the future within a preset time period are obtained.

[0021] The predicted rate of return is used as the predicted risk-reward result for users to select fund products.

[0022] The candidate fund product can be a single fund product or a fund portfolio consisting of multiple funds.

[0023] In one possible design, the method further includes:

[0024] Acquire sample data, which includes the target characteristics of multiple historical fund products and their corresponding historical actual returns;

[0025] Based on the sample data, continuous prediction is performed through a network model to obtain the predicted rate of return of historical fund products over a predefined historical period.

[0026] Based on the predicted rate of return and the corresponding historical actual rate of return, the parameters of the network model are adjusted by minimizing the loss function to obtain the product evaluation model.

[0027] In one possible design, determining the target fund product for recommendation from the plurality of candidate fund products based on the predicted risk-return outcome includes:

[0028] Determine whether the predicted risk-reward result is greater than or equal to the preset rate of return;

[0029] If the predicted risk-reward ratio is greater than or equal to the preset rate of return, then the candidate fund product is determined as the target fund product to be recommended to the user from among the multiple candidate fund products.

[0030] The target fund product includes at least one fund product.

[0031] Secondly, embodiments of this application provide a product recommendation device, the device comprising:

[0032] The acquisition module is used to acquire the target features of multiple candidate fund products. The target features include indicator features used to evaluate candidate fund products in the historical stage and personalized features of recommenders who recommended candidate fund products in the historical stage.

[0033] The prediction module is used to predict the risk and return outcomes of each of the candidate fund products based on the target characteristics and through a product evaluation model.

[0034] The product recommendation module is used to determine the target fund product for recommendation from the multiple candidate fund products based on the predicted return and risk results.

[0035] Thirdly, embodiments of this application provide an electronic device, including: at least one processor and a memory;

[0036] The memory stores computer-executed instructions;

[0037] The at least one processor executes computer execution instructions stored in the memory, causing the at least one processor to perform the product recommendation method as described in the first aspect and various possible designs of the first aspect.

[0038] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions. When a processor executes the computer-executable instructions, it implements the product recommendation method described in the first aspect and various possible designs of the first aspect.

[0039] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the product recommendation method described in the first aspect and various possible designs of the first aspect.

[0040] The product recommendation method, apparatus, device, storage medium, and program product provided in this embodiment first acquires target characteristics of multiple candidate fund products. These target characteristics include indicator characteristics used to evaluate candidate fund products in historical periods and personalized characteristics of recommenders who recommended candidate fund products in historical periods. Then, for each candidate fund product, based on the target characteristics, a product evaluation model is used to predict the risk-return outcome of each candidate fund product. Finally, based on the predicted risk-return outcome, a target fund product for recommendation is determined from the multiple candidate fund products. Therefore, by acquiring multi-dimensional characteristics of fund products and using a product evaluation model to assess or evaluate the risk-return of fund products, quantitative evaluation of fund products is achieved. The evaluation results (i.e., risk-return outcomes) are provided for recommenders to query and used as the basis for entering a preferred product pool, enabling accurate recommendation of target fund products to users. Attached Figure Description

[0041] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0042] Figure 1 A schematic diagram illustrating a scenario for the product recommendation method provided in this application embodiment;

[0043] Figure 2 A flowchart illustrating the product recommendation method provided in this application embodiment;

[0044] Figure 3 A flowchart illustrating a product recommendation method provided in another embodiment of this application;

[0045] Figure 4 A flowchart illustrating the LSTM model training process provided in this application embodiment;

[0046] Figure 5 A schematic diagram of the product recommendation device provided in the embodiments of this application;

[0047] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0049] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a particular order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the application described herein can be implemented, for example, in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0050] Currently, methods used to evaluate products include the Sharpe ratio, Treynor ratio, and Jensen alpha (also known as the Jensen index). However, the Sharpe ratio, based on the Capital Asset Pricing Model (CAPM) assumptions and using the risk-free rate as a benchmark, adjusts for total risk, thus leading to some bias in evaluating asset returns. The Treynor ratio, where a higher ratio indicates better asset performance, fails to consider the existence of unsystematic risk in actual fund portfolios, resulting in an overly idealistic measurement of risk. Jensen alpha, where a higher ratio indicates greater excess returns and better asset performance, and vice versa, does not reflect the actual excess returns achieved by either the market as a whole or individual assets. Therefore, current technologies cannot accurately evaluate fund products, and consequently, cannot accurately and effectively recommend fund products to users.

[0051] To address the problems existing in the prior art, the technical concept of this application is to obtain the multi-dimensional characteristics of fund products, use a product evaluation model to assess or evaluate the returns and risks of fund products, thereby achieving quantitative evaluation of fund products. The evaluation results (i.e., return and risk results) are provided for recommenders to query, and the evaluation results are used as the basis for entering the preferred product pool, which can accurately recommend target fund products to users.

[0052] In practical applications, see Figure 1 As shown, Figure 1 This is a schematic diagram illustrating a scenario for the product recommendation method provided in this application. The executing entity of this application can be a product recommendation device, which can be deployed in an electronic device, such as a terminal device or a server. The following detailed description uses an example where the client of the product recommendation device is installed on a terminal device, and the product recommendation process is executed by the server (e.g., a server) of the product recommendation device.

[0053] For example, a user can consult a recommender about fund products. The recommender can select multiple candidate fund products that meet the user's requirements based on the user's needs. Alternatively, if the user has no immediate needs, they can directly input all existing fund products as multiple candidate fund products into the client of the product recommendation device on the terminal device to trigger a product recommendation request. The server can first obtain the target features of multiple candidate fund products based on the input information. Then, it can input the indicator features used to evaluate candidate fund products in the historical period and the personalized features of the recommender who recommended candidate fund products in the historical period into the product evaluation model, and output the predicted return and risk results corresponding to each candidate fund product, such as the rate of return considering risk.

[0054] Then, based on the predicted risk and return results, fund products with higher returns are selected and stored in the preferred product pool to more accurately recommend fund products (which may refer to target fund products) to users. At the same time, based on user needs, products that meet personalized needs can be selected from the preferred product pool, thereby improving user satisfaction.

[0055] It should be noted that the collection, storage, use, processing, transmission, provision, and disclosure of financial data or user data involved in the technical solution of this application all comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0056] The technical solutions of this application will be described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0057] Figure 2 This is a flowchart illustrating a product recommendation method provided in an embodiment of this application. The method may include:

[0058] S201. Obtain the target characteristics of multiple candidate fund products.

[0059] The target features include the indicator features used to evaluate candidate fund products in the historical period and the personalized features of the recommenders who recommended candidate fund products in the historical period.

[0060] In this embodiment, the product recommendation method can be executed by a product recommendation device, which can be deployed in an electronic device, such as a terminal device or a server. Specifically, authorized queries can be made to retrieve product evaluation data, which is used to conduct an objective evaluation from the perspective of the fund's own returns and risks. Authorized queries can also be made to retrieve the personalized data of the recommender, which is used to evaluate the product based on the recommender's relevance to the fund selection. Then, based on the acquired data, the main features, i.e., the target features, are extracted to achieve dimensionality reduction.

[0061] S202. For each of the candidate fund products, based on the target characteristics, predict the return and risk outcomes of each of the candidate fund products using a product evaluation model.

[0062] In this embodiment, a network model can be trained, such as a Long Short-Term Memory (LSTM) model, which supports multiple inputs and multiple outputs. This means that the indicator features used to evaluate candidate fund products in the historical stage and the personalized features of the recommenders of candidate fund products in the historical stage are input into the trained LSTM model, i.e., the product evaluation model, to predict the return and risk results.

[0063] S203. Based on the predicted return and risk results, determine the target fund product for recommendation from the multiple candidate fund products.

[0064] In this embodiment, based on the predicted risk-return results, such as the rate of return, for each candidate fund product, fund products with higher rates of return are selected and stored in the preferred product pool to more accurately recommend fund products to users. At the same time, based on user needs, products that meet personalized needs can be selected from the preferred product pool, thereby improving user satisfaction.

[0065] The product recommendation method provided in this application obtains target characteristics of multiple candidate fund products. These target characteristics include indicator characteristics used to evaluate candidate fund products in historical periods and personalized characteristics of recommenders who recommended candidate fund products in historical periods. Then, for each candidate fund product, based on the target characteristics, a product evaluation model is used to predict the risk-return outcome of each candidate fund product. Finally, based on the predicted risk-return outcome, a target fund product for recommendation is determined from the multiple candidate fund products. Therefore, by obtaining multi-dimensional characteristics of fund products and using a product evaluation model to assess or evaluate the risk-return of fund products, quantitative evaluation of fund products is achieved. The evaluation results (i.e., risk-return outcomes) are provided for recommenders to query and used as the basis for entering a preferred product pool, enabling accurate recommendation of target fund products to users.

[0066] In one possible design, this embodiment provides a detailed description of S201 based on the above embodiment. The acquisition of the target features of multiple candidate fund products can be achieved through the following steps:

[0067] Step a1: Obtain feature-related data for multiple candidate fund products. The feature-related data includes indicator data used to evaluate candidate fund products in the historical period and personalized data of recommenders who recommended candidate fund products in the historical period. The indicator data includes at least one of the following: return rate indicator, Sharpe ratio, Treynor ratio, excess return, risk indicator, and investment type of fund product. The personalized data of recommenders includes at least one of the following: the recommender's basic personal data and data reflecting their capabilities.

[0068] Step a2: For the feature-related data of each candidate fund, extract the indicator features corresponding to the indicator data and the personalized features of the recommender corresponding to the personalized data of the recommender.

[0069] The target features consist of indicator features and the personalized features of the recommender.

[0070] In this embodiment, the target feature extraction process can be analyzed from two dimensions: the fund's own indicator characteristics and the relevance of the fund recommender's selection of the fund. Specifically, the indicator data may include:

[0071] Return rate indicators: Historical returns play a very important role in fund evaluation. In this embodiment, the returns over the past week, month, three months, six months, one year, three years, five years, cumulative returns since inception, and annualized returns over seven days since inception are taken as a major feature of the fund (here referring to a single fund).

[0072] Sharpe ratio: Represents the relationship between risk-reward and total risk, calculated as follows:

[0073] PI S =(ER i -r) / δ i

[0074] Among them, PI S ER is the Sharpe ratio. i Let r represent the investment return rate of fund i, r represent the risk-free rate, and δ represent the investment return rate of fund i. i This represents the standard deviation of the investment return rate of fund i.

[0075] Treynor Ratio: Similar to the Sharpe Ratio, but the Treynor Ratio uses the beta coefficient of the fund's investment returns as a measure of risk, instead of using standard deviation like the Sharpe Ratio. It is calculated as follows:

[0076] PI T =(ER i -r) / β i

[0077] Among them, PI T Represents the Treynor ratio, ER i Let r represent the investment return rate of fund i, r represent the risk-free rate, and β represent the return rate of fund i. i This represents the systematic risk of the investment return rate of fund i.

[0078] Jensen Alpha: Similar to the Treynor ratio, but this index uses another metric, the alpha value, calculated as follows:

[0079] α i =(ER i -r) / β i (R m -r)

[0080] Where, α i This refers to excess returns, also known as the Jensen's performance index; R m To assess the overall market return during the evaluation period; R m -r represents market risk compensation during the assessment period. When αi A significantly positive α value indicates that fund i performs better than the market; when α i A significantly negative value indicates that fund i's overall market performance is poor;

[0081] Risk indicators: They have a certain predictive role in the analysis of fund returns. In this example, annualized volatility, beta, tracking error, maximum drawdown, and lower standard deviation are used as risk indicators.

[0082] Basic indicators: The investment type of the product, namely money market fund, bond fund, mixed fund, and equity fund, is used as the basic indicator of the product.

[0083] Selection of fund recommender characteristic data:

[0084] Personal characteristics: Age, education, and investment experience were selected as personal characteristic indicators for evaluating fund recommenders.

[0085] Capability Characteristics: Market timing and stock selection abilities are selected as indicators to evaluate the investment capabilities of fund recommenders. The TM model assumes that fund recommenders can leverage their market timing ability to predict market returns. Fund recommenders with positive market timing ability will increase the portfolio beta when predicted market returns rise and decrease the portfolio beta when predicted market returns fall. The model formula is as follows:

[0086]

[0087] in, R is the arithmetic mean of the fund's adjusted net asset value growth rate series data. f R is the risk-free rate of return. m Let α represent the market return. The α obtained by regressing this model reflects the fund recommender's stock-picking ability; β can be used as the fund recommender's market timing ability.

[0088] Therefore, in feature selection, indicators from three classic product evaluation methods were incorporated, including the Sharpe ratio, Treynor ratio, and Jensen's alpha ratio as part of the data features. In addition, the recommender's personalized data was combined with the least squares method to calculate the recommender's stock selection and timing ability, and normalization was used to process the data, which can improve the accuracy of prediction.

[0089] In one possible design, extracting the indicator features corresponding to the indicator data and the personalized features of the recommender corresponding to the recommender's personalized data can be achieved through the following steps:

[0090] Step b1: Perform principal component regression analysis on the aforementioned return indicators, Sharpe ratio, Treynor ratio, excess return, and risk indicators to obtain the regression analysis results.

[0091] Step b2: Perform one-hot encoding on the investment type and the recommender's personal basic data to obtain encoded data;

[0092] Step b3: Process the capability data using the least squares method to obtain capability characteristics;

[0093] Step b4: Through the principal component regression analysis, the regression analysis results, coded data, and capability features are dimensionality reduced to obtain the target features.

[0094] In this embodiment, feature extraction is the data processing process. Common processing methods include one-hot encoding, feature scaling such as normalization and standardization, methods for constructing auxiliary models, regularization methods, and Principal Component Analysis (PCA). Among these, using principal component regression to process the data solves the problem of data collinearity.

[0095] For example, see Figure 3 As shown, Figure 3 This is a flowchart illustrating a product recommendation method provided in another embodiment of this application. Principal component regression analysis is performed on indicator data (such as returns, volatility, etc.). Least squares (normalization) is used to estimate the parameters of the fund recommender's characteristics. One-hot encoding is used for the product's basic characteristics, such as investment type and education level. Finally, PCA is applied to reduce the dimensionality of the data features, extract the main features (i.e., target features), and input the vector into the model.

[0096] In this embodiment, see Figure 4 The above, Figure 4 This is a flowchart illustrating the LSTM model training process provided in this application embodiment. After feature processing of the data, the features can be divided into two sets: the first 90% is used as the training set, and the last 10% is used as the test set. The LSTM model is then trained using the training set and validated using the test set. Dropout is set to 0.5, the Adam optimizer is used for optimization, the model runs for 100 epochs with a batch size of 100, and the loss function is evaluated and optimized using the root mean square error (RMSE).

[0097] Specifically, LSTM-based product yield prediction:

[0098] LSTM attempts to remember all the knowledge seen in the past and forget irrelevant data. It achieves different purposes through the forget gate function and is one of the best performing models in RNN. The internal structure of the hidden unit of LSTM is complex and mainly consists of three gate structures: forget gate, input gate, and output gate.

[0099] The hidden state is achieved through several mechanisms: Forget gate: determines the degree to which previous data has been forgotten, filtering out less important or unnecessary information; Input gate: determines the information to be written to the internal cell; Output gate: selects useful information and determines the output, i.e., the next hidden state.

[0100] By selectively forgetting or remembering the input information (which could refer to the extracted target features), the modeling ability for long-short-term dependencies is enhanced. The LSTM forward propagation is shown in the formula:

[0101]

[0102] Rolling forecast:

[0103] Rolling forecasts are continuous, based on a sliding window defined for the forecast. By sliding the window, the latest data is added to predict the value for the next period. For example, the value for day t is predicted using the values ​​of t-1, t-2, ..., t-30. Then, the data from day t is used as training data to predict the data for the next t+1, and so on. This yields the predicted data for the next 5 days. For example, the sliding window can be set to win = 30.

[0104] Model evaluation:

[0105] During training, the model is optimized by minimizing a loss function. Minimizing the loss function is equivalent to minimizing the squared difference between the predicted and actual values, where the predicted value is represented by... The true value is represented by y. t This indicates that an L2 regularization term was added to prevent overfitting, where λ is a hyperparameter, and its loss function is shown in the formula:

[0106]

[0107] Combination Figure 4 As shown, the feature data (here referring to the target features) is input into the LSTM model. When T=1, the LSTM model is trained, predicting the value on day t (i.e., predicting the return on day t) using the values ​​of t-1, t-2...t-30. Then, at T+1, it is determined whether T is greater than 5. If not, training continues. If yes, the model is evaluated, and it is determined whether the number of iterations has been reached. If yes, the model is saved and training ends, and the saved model is used to predict the return. If not, the sliding window is used to reset T=1, new feature data is added, and model training continues.

[0108] In one possible design, S202 is described in detail and can be implemented through the following steps:

[0109] Step d1: Input the target features into the product evaluation model, continuously predict the returns of each fund in the candidate fund products, and perform weighted calculation on the returns of each fund to obtain the predicted returns of the candidate fund products in the future preset time period.

[0110] Step d2: Use the predicted rate of return as the predicted return and risk result for users to select fund products.

[0111] The candidate fund product can be a single fund product or a fund portfolio consisting of multiple funds.

[0112] In this embodiment, LSTM is used to predict returns for product selection. This data-driven intelligent selection reduces bias caused by subjective choices. Specifically, target features are input into the product evaluation model, which continuously predicts the returns of each fund among the candidate funds. For example, the value for day t is predicted using the values ​​of t-1, t-2...t-30, and so on, until the value for day t+n is predicted. The returns of each fund are then weighted to obtain the predicted return of the candidate fund over a preset time period (each day from day t to day t+n).

[0113] In one possible design, determining the target fund product for recommendation from the plurality of candidate fund products based on the predicted risk-return outcome can be achieved through the following steps:

[0114] Determine whether the predicted risk-reward result is greater than or equal to the preset rate of return;

[0115] If the predicted risk-reward ratio is greater than or equal to the preset rate of return, then the candidate fund product is determined as the target fund product to be recommended to the user from among the multiple candidate fund products.

[0116] The target fund product includes at least one fund product.

[0117] In this embodiment, the predicted rate of return is compared with the preset rate of return. If the predicted rate of return is greater than or equal to the preset rate of return, it means that the fund product can be recommended to the user so that the user can make a more accurate product selection.

[0118] Therefore, this application proposes a time-series-based product selection method (where time-series analysis predicts future values ​​by observing historical data). In subsequent product management, the predicted product rate of return is used to evaluate the product, providing a more scientific and objective reference for product selection. Furthermore, since the product rate of return for the next five days is predicted each day, the selected products can be updated in real time based on the prediction results (which can refer to the predicted risk-reward outcome), thereby more accurately improving the supply capacity to customers.

[0119] To implement the product recommendation method, this embodiment provides a product recommendation device. See also... Figure 5 , Figure 5 This is a schematic diagram of the structure of a product recommendation device provided in an embodiment of this application; the product recommendation device 50 includes:

[0120] The acquisition module 501 is used to acquire the target features of multiple candidate fund products. The target features include indicator features used to evaluate candidate fund products in the historical stage and personalized features of recommenders who recommend candidate fund products in the historical stage.

[0121] The prediction module 502 is used to predict the return and risk outcomes of each of the candidate fund products based on the target characteristics and through a product evaluation model.

[0122] The product recommendation module 503 is used to determine the target fund product for recommendation from the multiple candidate fund products based on the predicted return and risk results.

[0123] In this embodiment, an acquisition module 501, a prediction module 502, and a product recommendation module 503 are configured to acquire target features of multiple candidate fund products. These target features include indicator features used to evaluate candidate fund products in historical periods and personalized features of recommenders who recommended candidate fund products in historical periods. Then, for each candidate fund product, based on the target features, a product evaluation model is used to predict the risk-return outcome of each candidate fund product. Finally, based on the predicted risk-return outcome, a target fund product for recommendation is determined from the multiple candidate fund products. Therefore, by acquiring multi-dimensional features of fund products and using a product evaluation model to assess or evaluate the risk-return of fund products, quantitative evaluation of fund products is achieved. The evaluation results (i.e., risk-return outcomes) are provided for recommenders to query and used as the basis for entering the preferred product pool, enabling accurate recommendation of target fund products to users.

[0124] The apparatus provided in this embodiment can be used to execute the technical solutions of the above method embodiments. Its implementation principle and technical effects are similar, and will not be described again here.

[0125] In one possible design, the acquisition module includes: a data acquisition unit and a feature extraction unit;

[0126] The data acquisition unit is used to acquire feature-related data of multiple candidate fund products. The feature-related data includes indicator data used to evaluate candidate fund products in historical periods and personalized data of recommenders who recommended candidate fund products in historical periods. The indicator data includes at least one of the following: return rate indicator, Sharpe ratio, Treynor ratio, excess return, risk indicator, and investment type of fund product. The personalized data of recommenders includes at least one of the following: basic personal data of recommenders and data reflecting their capabilities.

[0127] The feature extraction unit is used to extract the indicator features corresponding to the indicator data and the personalized features of the recommender corresponding to the personalized data of the recommender for each of the candidate funds' feature-related data.

[0128] The target features consist of indicator features and the personalized features of the recommender.

[0129] In one possible design, the feature extraction unit is specifically used for:

[0130] Principal component regression analysis was performed on the aforementioned return rate indicators, Sharpe ratio, Treynor ratio, excess return, and risk indicators to obtain the regression analysis results.

[0131] The investment type and the recommender's basic personal data are uniquely encoded to obtain encoded data;

[0132] The capability characteristics are obtained by processing the capability-reflecting data using the least squares method.

[0133] The principal component regression analysis is used to reduce the dimensionality of the regression analysis results, coded data, and capability features to obtain the target features.

[0134] In one design, the device further includes: a training module; the training module is used for:

[0135] Acquire sample data, which includes the target characteristics of multiple historical fund products and their corresponding historical actual returns;

[0136] Based on the sample data, continuous prediction is performed through a network model to obtain the predicted rate of return of historical fund products over a predefined historical period.

[0137] Based on the predicted rate of return and the corresponding historical actual rate of return, the parameters of the network model are adjusted by minimizing the loss function to obtain the product evaluation model.

[0138] In one possible design, the prediction module is specifically used for:

[0139] The target features are input into the product evaluation model. By continuously predicting the returns of each fund in the candidate fund products and weighting the returns of each fund, the predicted returns of the candidate fund products in the future within a preset time period are obtained.

[0140] The predicted rate of return is used as the predicted risk-reward result for users to select fund products.

[0141] The candidate fund product can be a single fund product or a fund portfolio consisting of multiple funds.

[0142] In one possible design, the product recommendation module is specifically used for:

[0143] Determine whether the predicted risk-reward result is greater than or equal to the preset rate of return;

[0144] When the predicted return risk is greater than or equal to the preset rate of return, the candidate fund product is determined as the target fund product to be recommended to the user from among the multiple candidate fund products;

[0145] The target fund product includes at least one fund product.

[0146] To implement the product recommendation method, this embodiment provides an electronic device. Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 6 As shown, the electronic device 60 of this embodiment includes: at least one processor 601 and a memory 602; wherein, the memory 602 is used to store computer execution instructions; and the at least one processor 601 is used to execute the computer execution instructions stored in the memory to implement the various steps performed in the above embodiment. For details, please refer to the relevant descriptions in the foregoing method embodiments.

[0147] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the product recommendation method described above.

[0148] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the product recommendation method described above.

[0149] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or modules, and may be electrical, mechanical, or other forms. Additionally, the functional modules in the various embodiments of this application may be integrated into one processing unit, or each module may exist physically separately, or two or more modules may be integrated into one unit. The above-mentioned modular units can be implemented in hardware or in the form of hardware plus software functional units.

[0150] The integrated modules implemented as software functional modules described above can be stored in a computer-readable storage medium. These software functional modules, stored in a storage medium, include several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute some steps of the methods described in the various embodiments of this application. It should be understood that the processor may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. A general-purpose processor may be a microprocessor or any conventional processor. The steps of the methods disclosed in this invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.

[0151] The memory may include high-speed RAM, and may also include non-volatile memory (NVM), such as at least one disk drive, and may also be a USB flash drive, external hard drive, read-only memory, disk, or optical disc. The bus may be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses in the accompanying drawings are not limited to a single bus or a single type of bus. The aforementioned storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, disk, or optical disc. The storage medium can be any available medium accessible to general-purpose or special-purpose computers.

[0152] An exemplary storage medium is coupled to a processor, enabling the processor to read information from and write information to the storage medium. Alternatively, the storage medium can be an integral part of the processor. Both the processor and the storage medium can reside in an Application Specific Integrated Circuit (ASIC). Alternatively, the processor and storage medium can exist as discrete components in an electronic device or host device.

[0153] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0154] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

Claims

1. A product recommendation method characterized by, The method comprises: obtaining target features of a plurality of candidate fund products, the target features comprising index features used for evaluating the candidate fund products in a historical stage and personalized features of recommenders who recommend the candidate fund products in the historical stage; for each of the candidate fund products, predicting a yield risk result of each of the candidate fund products according to the target features through a product evaluation model; determining a target fund product for recommendation from the plurality of candidate fund products according to the predicted yield risk result; the obtaining of the target features of the plurality of candidate fund products comprises: obtaining feature-related data of a plurality of candidate fund products, the feature-related data comprising index data used for evaluating the candidate fund products in a historical stage and personalized data of recommenders who recommend the candidate fund products in the historical stage; the index data comprises at least one of the following: yield rate index, Sharpe ratio, Treynor ratio, excess yield, risk index, investment type of the fund product, and the personalized data of the recommenders comprises at least one of the following: personal basic data and ability reflecting data of the recommenders; for each of the feature-related data of the candidate fund, the index data corresponding to the index features and the personalized data of the recommenders corresponding to the personalized features of the recommenders are extracted respectively; wherein the target features are composed of the index features and the personalized features of the recommenders; the method further comprises: obtaining sample data, the sample data comprising target features of a plurality of historical fund products and corresponding historical actual yield rates; performing continuous prediction through a network model according to the sample data to obtain predicted yield rates of the historical fund products in a pre-defined historical time period; adjusting parameters of the network model by minimizing a loss function according to the predicted yield rates and the corresponding historical actual yield rates to obtain a product evaluation model.

2. The method of claim 1, wherein, the extracting of the index features corresponding to the index data and the personalized features of the recommenders corresponding to the personalized data of the recommenders respectively comprises: performing principal component regression analysis on the yield rate index, Sharpe ratio, Treynor ratio, excess yield, risk index to obtain regression analysis results; performing one-hot encoding on the investment type and the personal basic data of the recommenders to obtain encoded data; processing the ability reflecting data through a least square method to obtain ability features; performing dimension reduction processing on the regression analysis results, the encoded data and the ability features through the principal component regression analysis to obtain target features.

3. The method of claim 1, wherein, the predicting of the yield risk result of each of the candidate fund products according to the target features through the product evaluation model comprises: inputting the target features into the product evaluation model, predicting yield rates of each of the candidate fund products through continuous prediction, and performing weighted calculation on the yield rates of each of the fund products to obtain predicted yield rates of the candidate fund products in a future pre-set time period; using the predicted yield rates as the predicted yield risk results for users to select fund products; wherein the candidate fund products are products of a single fund or products of a fund combination composed of a plurality of funds.

4. The method according to claim 1 or 2, characterized in that, The method comprises: determining whether the predicted yield risk result is greater than or equal to a preset yield rate; if the predicted yield risk is greater than or equal to the preset yield rate, determining the candidate fund product as a target fund product for recommendation to the user from the plurality of candidate fund products; wherein the target fund product comprises at least one fund product.

5. A product recommendation device characterized by comprising: The device comprises: an acquisition module configured to acquire target features of a plurality of candidate fund products, the target features comprising index features used to evaluate the candidate fund products in a historical stage and personalized features of recommenders who recommend the candidate fund products in the historical stage; a prediction module configured to, for each of the candidate fund products, predict a yield risk result of each of the candidate fund products according to the target features through a product evaluation model; a product recommendation module configured to determine a target fund product for recommendation from the plurality of candidate fund products according to the predicted yield risk result. The acquisition module is specifically configured to: acquire feature-related data of the plurality of candidate fund products, the feature-related data comprising index data used to evaluate the candidate fund products in a historical stage and personalized data of recommenders who recommend the candidate fund products in the historical stage; the index data comprises at least one of the following: a yield rate index, a Sharpe ratio, a Treynor ratio, excess yield, a risk index, and an investment type of the fund product, and the personalized data of the recommenders comprises at least one of the following: personal basic data and ability reflecting data of the recommenders; for each of the candidate fund products, extract index features corresponding to the index data and personalized features of the recommenders corresponding to the personalized data of the recommenders from the feature-related data of the candidate fund products; wherein the target features are composed of the index features and the personalized features of the recommenders; The device further comprises a training module. The training module is specifically configured to: acquire sample data, the sample data comprising target features of a plurality of historical fund products and corresponding historical actual yield rates; perform continuous prediction through a network model according to the sample data to obtain predicted yield rates of the historical fund products in a predefined historical time period; adjust parameters of the network model by minimizing a loss function according to the predicted yield rates and the corresponding historical actual yield rates to obtain the product evaluation model.

6. An electronic device, comprising: The device comprises: at least one processor and a memory; the memory stores computer execution instructions; the at least one processor executes the computer execution instructions stored in the memory, so that the at least one processor performs the product recommendation method according to any one of claims 1 to 4.

7. A computer readable storage medium characterized in that, The computer readable storage medium stores computer execution instructions, and when the processor executes the computer execution instructions, the product recommendation method according to any one of claims 1 to 4 is implemented.

8. A computer program product comprising a computer program, characterized in that, The computer program is executed by the processor to implement the product recommendation method according to any one of claims 1 to 4.

Images