傳統的保險承保與精算中，一直以來都不斷在搜尋有意義的資訊指標，以明確依風險特徵進行對應訂價，保險公司對於客戶的風險評估，一般是基於要保人所填寫要保書中所揭露基本資料、健康告知與財力說明等資料，有關基本資料中【年齡】、【性別】、【婚姻】等基本資料，係為明確的數字或狀態(男或女、已婚或單身)，可以協助保險公司認識客戶大約樣貌，但如要更清楚 瞭解客戶輪廓，就需要【職業(服務公司)】、【職稱】等更多資料，來協助公司進一步辨識。目前個案公司在承接其客戶投保申請時，需透過客戶提供的各項資料進一步分析核保的項目與核保風險等。其中一項重要的依據是客戶的職業型態，個案公司依據客戶所服務的單位配對到個案公司內部的職業類別，進而評估核保風險，然而現階段在客戶服務單位與職業類別之配對仍採取人工方式進行判讀。人工判讀除了消耗大量人力與工時之外，同時亦依賴核保人員專業經驗，且職業配對之決策也會因人、因時而異；此外人工判讀無法快速應付壽險業龐大客戶資料之辨識。本研究計畫將專注開發【職業類別智能檢索系統】以解決個案公司大量人力判讀成本。運用大數據資料可為保險公司運作上帶來效益，然而，如何將資料探勘內容分析結果在實務上可行，並與傳統數據庫中的結構化資料兩者相結合，並強化改善決策制定，是保險公司所要面對的挑戰之一，因此本計畫主要目的是如何有效運用機器學習(資料探勘) 、自然語言處理(包含統計與符號或語義) 、以及人工智能(智能系統)等多元科技技術，開發一套可以解決保險產業在風險分類上之智能系統，並以關鍵核心風險指標職業做為智能系統之主軸。 ;In conventional insurance underwriting and actuarial calculations, insurers have been searching for meaningful indicators that help underwriters to give appropriate quotations and to carry out risk assessment. The insurance company's risk assessment for customers is generally based on the insurance applicants filling out the required documents. Basic personal details such as age, gender, and marital status provide some rough profile of the customers. To help the insurance company have a better picture of their customers, details such as occupation and job type are required.At present, when an underwriter receives an insurance application, he or she needs to manually look up the company for which the applicant works and then further extract the customer's job type in order to fully consider the application. This process consumes huge amount of human labor, and often to result in discrimination in application decision between different underwriters. The insurance company is looking for a better solution that could help reduce the time taken and the errors occurred in this process. This project will focus on the development of an Adaptive Intelligent Occupation Mapping System that will employ state-of-the-art machine learning and natural language processing techniques. The project team expects the system to help speed up the underwriting process with better accuracy and hopes to develop a standard automated workflow that will support insurance companies in risk assessment and insurance underwriting.
