Method for constructing intelligent recommendation platform based on artificial intelligence portrait system

Abstract

The invention discloses a method for constructing an intelligent recommendation platform based on an artificial intelligence portrait system. The method is specifically implemented according to the following steps: step 1, constructing a customer portrait system; 2, constructing a financial product label portrait system; 3, matching the customer portrait system with the financial product label portrait system; and step 4, realizing accurate customer screening by adopting an intelligent recommendation technology, and completing financial product recommendation. According to the method, a more accurate customer portrait and financial product label portrait system is constructed, personalized financial products and services are provided for customers, and the problems that an existing recommendation platform construction method is low in efficiency and poor in accuracy are solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of big data processing, and specifically relates to a method for building an intelligent recommendation platform based on an artificial intelligence portrait system. Background Art

[0002] With the vigorous development of the digital economy, data, computing power, and algorithms are fully integrated and play a synergistic innovation effect, promoting the in-depth application of artificial intelligence in the financial field, supporting the optimization of financial service models, innovating financial products, improving risk intelligence control, improving the quality and efficiency of financial services, and fully empowering financial institutions to transform digitally and digitally. Customers' financial needs are becoming increasingly diversified, and how to provide personalized financial services to tens of millions of retail customers has become a new issue in the financial field.

[0003] The Chinese patent "A method and system for intelligent recommendation based on customer portrait" (application date: 2021.09.26; application number: CN202111131023.2; announcement date: 2023.08.25; publication number: CN113836379A) ​​discloses a method and system for intelligent recommendation based on customer portrait, which obtains multiple customer names input by the user; crawls the information corresponding to each of the multiple customer names from the network; extracts the keywords corresponding to the customer information corresponding to each customer name according to the information corresponding to each customer name; obtains the same keywords corresponding to multiple customer names; obtains other customer names corresponding to the same keywords; through the above steps, the correspondence between the pre-stored keywords and the multiple customer names is obtained, and when the user retrieves the customer name in the corresponding relationship, the retrieved other customer names are displayed to the user, so as to achieve the effect of intelligent recommendation.

[0004] The above customer portrait construction method has the following problems:

[0005] 1. Lack of comprehensiveness and accuracy. The construction of customer portraits relies only on limited customer data and labels the data based on a small number of dimensions, which makes it difficult to provide a comprehensive and in-depth description of customers. There are certain errors in the data analysis and processing process, which reduces the accuracy of customer portraits and affects the ability to make accurate marketing decisions.

[0006] Second, lack of efficiency and real-time performance: Customer portrait construction requires analysis and processing of a large amount of customer data to extract customer characteristics and behavioral patterns. Manual data cleaning, conversion, and organization are labor-intensive and error-prone, and the processing efficiency for large-scale data sets is low. At the same time, it is also impossible to capture and analyze customer changes and trends in a timely manner, and conduct timely analysis and prediction, resulting in information lag and distortion.

[0007] 3. Lack of personalization: Personalized recommendations need to integrate multiple dimensions to better meet customer needs and expectations. Since the data processing dimension is too single, the customer portrait of a single dimension tends to be homogenized and poorly generalized, making it difficult to achieve personalized services and marketing, and unable to provide accurate personalized recommendations and customized experiences. Summary of the invention

[0008] The purpose of the present invention is to provide a method for building an intelligent recommendation platform based on an artificial intelligence portrait system, which solves the problems of low efficiency and poor accuracy in the existing methods for building recommendation platforms.

[0009] The technical solution adopted by the present invention is a method for building an intelligent recommendation platform based on an artificial intelligence portrait system, which is specifically implemented according to the following steps:

[0010] Step 1: Build a customer portrait system;

[0011] Step 2: Build a financial product labeling system;

[0012] Step 3: Match the customer portrait system with the financial product label portrait system to achieve accurate customer screening and complete financial product recommendations.

[0013] The present invention is also characterized in that:

[0014] The specific process of step 1 is:

[0015] Step 1.1, collect financial data of corporate and institutional customers, pre-process the customer financial data, and store the pre-processed customer financial data in the data warehouse raw data;

[0016] Step 1.2: Use the decision tree model to continuously train the model on the original data of the data warehouse, build a training model, and obtain feature classification and customer feature prediction values ​​according to the degree of contribution;

[0017] Step 1.3, adjust model parameters and optimize the model;

[0018] Step 1.4: Deploy and apply the model in step 1.3 online, generate customer characteristics, behaviors, preferences and other profile information according to the parameter tuning rules, use AI deep learning algorithms to automatically extract and identify corporate and institutional customer profile features, and automatically build a customer profile system.

[0019] The specific process of step 2 is:

[0020] Step 2.1, collect financial product data and perform preprocessing;

[0021] Step 2.2, classify the financial product data according to nature, purpose, and risk characteristic attributes;

[0022] Step 2.3: Extract and filter the characteristic information that has a greater impact on financial products to form a financial product portrait label system;

[0023] Step 2.4, use rule engine technology to achieve dynamic management of the financial product label portrait system, and adjust and optimize labels at any time according to customer needs and market changes.

[0024] The specific process of step 3 is:

[0025] Step 3.1, design a financial product portfolio based on customer needs and risk tolerance, combining products with different risk levels, yields and liquidity to achieve risk diversification and return optimization;

[0026] Step 3.2, based on the customer's personal characteristics and investment preferences, a content-based recommendation algorithm is used to match financial products with similar characteristics using the customer's historical behavior and existing preferences.

[0027] Step 3.3, using the cosine similarity algorithm or the Euclidean distance or the Pearson correlation coefficient method to calculate customer similarity, and calculate the similar customer group for each target customer;

[0028] Step 3.4, generate a recommendation list of smart financial products based on personalized recommendations for target customers and personalized recommendations for similar customer groups.

[0029] The beneficial effects of the present invention are as follows: the method of constructing an intelligent recommendation platform based on an artificial intelligence portrait system has the advantages of high efficiency, automation, precision and dynamism. The AI ​​artificial intelligence deep learning algorithm is used to automatically extract and identify the characteristics of corporate and institutional customer portraits, and automatically construct a customer portrait system; intelligent recommendation technology is used to achieve accurate customer screening, improve sales conversion rate, and optimize customer experience; the present invention uses rule engine technology to achieve dynamic management of the financial product label portrait system, and can adjust and optimize labels at any time according to customer needs and market changes, effectively improving the efficiency and quality of financial product sales and customer service, and has high practical value and promotion and application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 This is a flowchart of step 3 of the method for building an intelligent recommendation platform based on an artificial intelligence portrait system of the present invention;

[0031] Figure 2 This is a functional architecture diagram of the intelligent recommendation platform built based on the artificial intelligence portrait system in the present invention. DETAILED DESCRIPTION

[0032] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0033] Example 1

[0034] The method of constructing an intelligent recommendation platform based on an artificial intelligence portrait system of the present invention is specifically implemented according to the following steps:

[0035] Step 1: Build a customer portrait system;

[0036] Step 2: Build a financial product labeling system;

[0037] Step 3: Match the customer portrait system with the financial product label portrait system to achieve accurate customer screening and complete financial product recommendations.

[0038] Example 2

[0039] The method of constructing an intelligent recommendation platform based on an artificial intelligence portrait system of the present invention is specifically implemented according to the following steps:

[0040] Step 1: Build a customer portrait system;

[0041] Step 1.1, collect customer financial data, pre-process the customer financial data, and store the pre-processed customer financial data in the data warehouse raw data; pre-processing includes cleaning and deduplication operations on the collected corporate and institutional customer financial data.

[0042] Step 1.2, build a training model; use the decision tree model to continuously train the model on the original data of the data warehouse, build a training model, and obtain feature classification and customer feature prediction values ​​according to the degree of contribution;

[0043] Step 1.3, adjust model parameters and optimize the model;

[0044] Step 1.4: Deploy and apply the model in step 1.3 online, and generate customer characteristics, behavior, and preference profile information according to the parameter tuning rules; use AI artificial intelligence deep learning algorithms to automatically extract and identify corporate and institutional customer profile features, and automatically build a customer profile system.

[0045] Step 1: Collect financial data of corporate and institutional customers through the data center, and perform pre-processing operations such as cleaning and deduplication on the data. The collected data is stored in the data warehouse; use the decision tree model to continuously train the model on the original data in the data warehouse, build the model, and derive feature classification and customer feature prediction values ​​according to the degree of contribution; adjust and optimize model parameters to improve model accuracy; after the model is deployed and applied online, generate customer portraits, and generate portrait information such as customer characteristics, behaviors, and preferences based on the online model and original data; build a customer portrait system by clarifying needs and goals, and determine the scope of the customer portrait to be constructed.

[0046] Step 2: Build a financial product labeling system;

[0047] Step 2.1, collect financial product data and perform preprocessing;

[0048] The collection of financial product data includes basic information, product features, sales channels, and applicable population characteristics of each type of financial product;

[0049] Step 2.2, classify the financial product data according to nature, purpose, and risk characteristic attributes;

[0050] Step 2.3: Extract and filter the characteristic information that has a greater impact on financial products to form a financial product portrait label system;

[0051] Step 2.4, use rule engine technology to achieve dynamic management of the financial product label portrait system, and adjust and optimize labels at any time according to customer needs and market changes.

[0052] Step 2: Classify and summarize financial products by nature, purpose, risk and other characteristics; collect financial product characteristics, collect basic information, product features, sales channels, applicable population and other characteristic information of each type of financial product; select financial product characteristics, extract and screen characteristic information that has a greater impact on financial products, generate a financial product label portrait system, form a financial product label portrait based on the selected financial product characteristics, and construct a financial product portrait label system. By clarifying the goals and usage scenarios, the purpose of constructing financial products and the scope of specific application scenarios are clearly defined.

[0053] Step 3: Match the customer portrait system with the financial product label portrait system to achieve accurate customer screening and complete financial product recommendations.

[0054] like Figure 1 As shown in step 3.1, design a financial product portfolio based on customer needs and risk tolerance, combining products with different risk levels, yields and liquidity to achieve risk diversification and return optimization;

[0055] Step 3.2, based on the customer's personal characteristics and investment preferences, a content-based recommendation algorithm is used to match financial products with similar characteristics using the customer's historical behavior and existing preferences.

[0056] Step 3.3, calculate customer similarity and calculate the similar customer groups for each target customer;

[0057] Step 3.4: Generate a recommendation list of intelligent financial products based on personalized recommendations for target customers and similar customer groups, thereby improving sales conversion rate and optimizing customer experience.

[0058] Step 3: Through financial product portfolio design, product portfolios are designed according to customer needs and risk tolerance, and products with different risk levels, yields and liquidity are combined to achieve risk diversification and return optimization; personalized recommendations are made based on customer personal characteristics and investment preferences, using content-based recommendation algorithms, using customers' historical behaviors and existing preferences to match financial products with similar characteristics, and then using the portrait characteristics and needs of the customer portrait system to match customers' preferences and risk preferences with suitable financial products.

[0059] Example 3

[0060] The method of constructing an intelligent recommendation platform based on an artificial intelligence portrait system of the present invention is specifically implemented according to the following steps:

[0061] Step 1: Build a customer portrait system;

[0062] Step 1.1, collect customer financial data, including age, gender, occupation, income, investment history, family situation, etc.; pre-process the customer financial data and store the pre-processed customer financial data in the data warehouse raw data; pre-processing includes cleaning and deduplication of the collected corporate and institutional customer financial data.

[0063] Data cleaning includes cleaning customer financial data, removing duplicate data, processing missing values, standardizing data, processing outliers, and balancing data.

[0064] Step 1.2, build a training model; use the decision tree model to continuously train the model on the original data of the data warehouse, build a training model, and obtain feature classification and customer feature prediction values ​​according to the degree of contribution;

[0065] Step 1.3, adjust model parameters and optimize the model; adjust the model parameters through the gradient descent optimization algorithm so that the model can better fit the training data; use the evaluation indicators of accuracy, precision, and recall to evaluate the performance of the model on the test data set. According to the evaluation results, adjust the model's hyperparameters or improve the model structure.

[0066] Step 1.4: deploy the model in step 1.3 and put the application online. Deploy the trained model to actual applications, provide external interfaces through API or other means, and allow customers to obtain the model's prediction results by inputting data; optimize and improve the model based on feedback and needs in actual use, and improve the model's performance and effect by adjusting the model structure, increasing training data, and introducing new features.

[0067] Generate customer behavior characteristics, risk preferences, and investment purpose profile information according to parameter tuning rules; use AI artificial intelligence deep learning algorithms to automatically extract and identify corporate and institutional customer profile characteristics and automatically build a customer profile system. Figure 2 As shown, the functional architecture includes data layer, service layer and application layer.

[0068] The data layer includes the collection of financial data of corporate and institutional clients, including customer data, product data, behavioral data, and external data.

[0069] The service layer includes the data middle platform, AI middle platform, label middle platform, and public services; the data middle platform pre-processes the financial data of enterprise and institutional customers at the data layer, deduplicates, cleans, extracts, and performs data governance operations; the AI ​​middle platform continuously trains models on the original data in the warehouse, optimizes model parameters, and uses deep learning for prediction and recommendation; the label middle platform maintains basic label information and uses the AI ​​middle platform to label label data to customer entities and financial product entities; the public service monitors the system's log behavior and performs distributed task scheduling.

[0070] The application scenarios of the application layer include three dimensions: precision marketing, intelligent recommendation, and business opportunity matching.

[0071] AI deep learning algorithms continuously observe, experiment, and improve through an iterative process. Based on actual problems and changes in user needs, they adjust and optimize the models in a timely manner to meet changing needs, thereby effectively building a customer portrait system and providing customers with more personalized and accurate financial services.

[0072] Step 2: Build a financial product labeling system;

[0073] Step 2.1, collect financial product data and perform preprocessing;

[0074] The collection of financial product data includes basic information, product features, sales channels, and applicable population characteristics of each type of financial product;

[0075] Step 2.2, classify the financial product data according to risk characteristics, return characteristics, and investment strategy characteristics;

[0076] Step 2.3: Extract and filter the characteristic information that has a greater impact on financial products to form a financial product portrait label system;

[0077] Step 2.4, use rule engine technology to achieve dynamic management of the financial product label portrait system, and adjust and optimize labels at any time according to customer needs and market changes.

[0078] Step 3: Match the customer portrait system with the financial product label portrait system to achieve accurate customer screening and complete financial product recommendations.

[0079] Step 3.1, design a financial product portfolio based on customer needs and risk tolerance, combining products with different risk levels, yields and liquidity to achieve risk diversification and return optimization; the combination includes data transformation, feature engineering, data splitting, data normalization, data enhancement and data set management, etc.

[0080] Step 3.2, based on the customer's personal characteristics and investment preferences, use content-based recommendation algorithms to build product recommendation models and risk assessment models, and use the customer's historical behavior and existing preferences to match financial products with similar characteristics.

[0081] Step 3.3, calculate customer similarity and calculate the similar customer groups for each target customer;

[0082] The method for calculating customer similarity uses the cosine similarity algorithm or the Euclidean distance or the Pearson correlation coefficient method to calculate the similarity between customers. In practical applications, customer portraits often involve multiple features, each of which may have different weights. When calculating similarity, different features are given corresponding weights to better reflect the similarity or correlation between customers.

[0083] Step 3.4: Generate a recommendation list of intelligent financial products based on personalized recommendations for target customers and similar customer groups, thereby improving sales conversion rate and optimizing customer experience.

[0084] The present invention is a method for constructing an intelligent recommendation platform based on an artificial intelligence portrait system, which uses an AI artificial intelligence deep learning algorithm to automatically extract and identify the characteristics of corporate and institutional customer portraits, and automatically construct a customer portrait system. The rule engine technology is used to realize the dynamic management of the financial product label portrait system, and the labels can be adjusted and optimized at any time according to customer needs and market changes. Intelligent recommendation technology is used to realize accurate customer screening, improve sales conversion rate, and optimize customer experience. Automatically extracting and identifying corporate and institutional customer portrait characteristics, and automatically building a customer portrait system can greatly improve the accuracy and completeness of customer portraits, and support more accurate customer positioning recommendation services. Dynamic management of the financial product label portrait system can adjust and optimize labels according to customer needs and market changes, support more flexible and intelligent product recommendations and sales; realize accurate customer screening, improve sales conversion rate, optimize customer experience, and support more efficient and personalized sales services and customer relationship management.

Claims

1. A method for building an intelligent recommendation platform based on an artificial intelligence portrait system, characterized in that: Follow the steps below to implement it: Step 1: Build a customer portrait system; Step 2: Build a financial product labeling system; Step 3: Match the customer portrait system with the financial product label portrait system to achieve accurate customer screening and complete financial product recommendations.

2. The method for building an intelligent recommendation platform based on an artificial intelligence portrait system according to claim 1, characterized in that: The specific process of step 1 is: Step 1.1, collect customer financial data, pre-process the customer financial data, and store the pre-processed customer financial data in the data warehouse raw data; Step 1.2, build a training model; Step 1.3, adjust model parameters and optimize the model; Step 1.4: Deploy and apply the model in step 1.3 online, generate customer characteristics, behaviors, preferences and other profile information according to the parameter tuning rules, and automatically build a customer profile system.

3. The method for building an intelligent recommendation platform based on an artificial intelligence portrait system according to claim 2, characterized in that: The specific process of step 2 is: Step 2.1, collect financial product data and perform preprocessing; Step 2.2, classify the financial product data according to nature, purpose, and risk characteristic attributes; Step 2.3: Extract and filter the characteristic information that has a greater impact on financial products to form a financial product portrait label system; Step 2.4, use rule engine technology to achieve dynamic management of the financial product label portrait system, and adjust and optimize labels at any time according to customer needs and market changes.

4. The method for building an intelligent recommendation platform based on an artificial intelligence portrait system according to claim 3, characterized in that: The specific process of step 3 is: Step 3.1, design a financial product portfolio based on customer needs and risk tolerance, combining products with different risk levels, yields and liquidity to achieve risk diversification and return optimization; Step 3.2, based on the customer's personal characteristics and investment preferences, a content-based recommendation algorithm is used to match financial products with similar characteristics using the customer's historical behavior and existing preferences; Step 3.3, calculate customer similarity and calculate the similar customer groups for each target customer; Step 3.4: Generate a recommendation list of smart financial products based on personalized recommendations for target customers and similar customer groups.

5. The method for building an intelligent recommendation platform based on an artificial intelligence portrait system according to claim 2, characterized in that: The preprocessing in step 1.1 includes cleaning and deduplication of the collected financial data of corporate and institutional clients.

6. The method for building an intelligent recommendation platform based on an artificial intelligence portrait system according to claim 2, characterized in that: In step 1.2, the decision tree model is used to continuously train the original data in the data warehouse, build a training model, and obtain feature classification and customer feature prediction values ​​according to the degree of contribution.

7. The method for building an intelligent recommendation platform based on an artificial intelligence portrait system according to claim 2, characterized in that: In step 1.4, AI deep learning algorithms are used to automatically extract and identify the characteristics of corporate and institutional customer portraits.

8. The method for building an intelligent recommendation platform based on an artificial intelligence portrait system according to claim 4, characterized in that: Step 3.3 uses the cosine similarity algorithm or the Euclidean distance or the Pearson correlation coefficient method to calculate customer similarity.

Images