Ping An Insurance (China) applies for a patent on an AI-powered method for identifying high-value, high-risk attrition agents: Boosting precision marketing and risk management

Ping An Insurance (China) applies for a patent on an AI-powered method for identifying high-value, high-risk attrition agents: Boosting precision marketing and risk managementThe State Intellectual Property Office of China shows that Ping An Life Insurance Company of China Limited filed a patent application in September 2024 for an "AI-based population identification method, device, computer equipment and medium," with publication number CN119312194A. This patent falls under the intersection of artificial intelligence and fintech, aiming to leverage AI to more accurately identify high-value, high-risk attrition agents, thus providing insurance companies with more effective marketing strategies and risk management solutions

Ping An Insurance (China) applies for a patent on an AI-powered method for identifying high-value, high-risk attrition agents: Boosting precision marketing and risk management

The State Intellectual Property Office of China shows that Ping An Life Insurance Company of China Limited filed a patent application in September 2024 for an "AI-based population identification method, device, computer equipment and medium," with publication number CN119312194A. This patent falls under the intersection of artificial intelligence and fintech, aiming to leverage AI to more accurately identify high-value, high-risk attrition agents, thus providing insurance companies with more effective marketing strategies and risk management solutions.

The core of this patent application lies in an AI-based population identification method. This method begins by acquiring agent profile data encompassing three dimensions: basic attributes, value attributes, and behavioral attributes. Basic attributes may include demographic information such as age, gender, and region; value attributes may include performance metrics like sales, client count, and premium volume; and behavioral attributes may record work frequency, client communication methods, and business processing efficiency, reflecting work style and patterns.

After acquiring complete profile data, the method undergoes several data processing steps. First, the raw profile data undergoes feature engineering to extract features more effective for subsequent analysis, resulting in the first profile data set. This step is crucial, determining the accuracy and efficiency of subsequent clustering and classification models. Feature engineering may involve data cleaning, transformation, selection, and combination.

Next, the first profile data set is standardized to produce the second data set. Standardization aims to eliminate the influence of different feature scales, ensuring comparability and guaranteeing the fairness and effectiveness of clustering and classification algorithms. Common standardization methods include z-score and min-max normalization.

Following standardization, the method employs clustering algorithms to analyze the second profile data set. The goal is to group agents with similar characteristics into the same category. Common clustering algorithms include K-Means, DBSCAN, and hierarchical clustering. Clustering identifies high-value groups from all agents.

After identifying high-value groups, the method uses a pre-set target classification model to generate attrition probabilities for these individuals. This classification model may be a machine learning prediction model such as logistic regression, support vector machines, or neural networks. Training data may include historical agent data, containing attributes, behaviors, and attrition labels. This allows the model to learn attrition patterns and predict future attrition probabilities for high-value agents.

Finally, the method filters the high-value group based on calculated attrition probabilities to identify high-value, high-risk attrition agents. These individuals require focused attention and measures to mitigate attrition risk, such as competitive compensation, improved management, and enhanced training. Agents with high attrition probabilities are prioritized for intervention.

Notably, the patent application mentions storing attrition probabilities on a blockchain. This suggests secure and reliable storage and management of agent attrition probability information, enhancing data transparency and traceability, and ensuring data security and integrity. Blockchain technology effectively prevents data tampering and loss, providing more reliable risk management.

This patent application's technical solution, combining clustering algorithms and target classification models, enables automatic and precise identification of high-value, high-risk attrition agents. This significantly improves the efficiency of insurance companies' precision marketing and risk management, helping retain key talent, reduce operating costs, and increase profitability.

Ping An Life Insurance Company of China Limited was established in 2002, is located in Shenzhen, has registered capital of RMB 33,800 million, and paid-in capital of RMB 33,635,361,364. Tianyancha data shows that the company has invested in 168 companies, participated in 133 bidding projects, holds 3580 patents, and has 72 administrative licenses, demonstrating its substantial strength and influence. This patent application is further evidence of Ping An's continuous innovation in AI and fintech, signifying new development opportunities for the insurance industry's digital transformation and intelligent upgrade. The application of this technology will not only improve internal management but also contribute to the overall advancement of the industry. Precise identification of high-value, high-risk attrition groups allows insurance companies to allocate resources and adjust strategies more effectively, ultimately maximizing efficiency. This will undoubtedly promote the healthy development of the entire insurance industry and provide customers with better services.