結合社群資料與現有信用評估參考指標進行信用分類

健全的經濟環境仰賴於有效率的金融體制。而對於銀行內來說，最重要的活動莫過於信用風險管理;銀行運用信用評分來依風險等級區分潛在客戶，提供不同等級的監控與服務。許多企業正視到過去評分模型不足之處，開始從資料來源著手，期望改變整個評分生態，並期望從中把握機會獲利。銀行在這場信用評分環境的變革下，也在漸漸調整舊有參考項目。此篇論文站在銀行的角度，思考未來社群資料對信用評分的影響與發展，本研究將不會處理個人評分，而是聚焦於整個顧客群。藉由過去銀行於信用評分上所採用之參考項目，結合個人在社群媒體上的好友關係，來為整個顧客群做信用分類。本研究希望從實驗結果一窺目前各大銀行的未來行動方針，並提供他們可行的信評解決方案。主要研究問題如下： (一) 社群在日常生活中的影響力多大?是否值得列為信用評分的參考變數? (二) 探討涵蓋社群資料之信用分類是否對信用評估與風險控管有直接或間接的影響?能否改變過去只看個人財經歷史紀錄所造成的缺失? (三) 探討包含社群資料之新的信用評分方式是否能從現有顧客中找出潛在關聯，並藉此獲利? (四) 探討包含社群資料之新的信用評分方式在未來的發展性與應用層面，在當下適合用於判斷哪些行業的人?未來又能涵蓋到哪些領域?又在什麼情境下使用最合宜? 並在研究最後，提出未來可以改善的方向與應用領域，期望帶給銀行一個更有效的信用評估方式。

關鍵字

信用評分；信用分數；社群媒體；支持向量機；類神經網路；邏輯回歸

並列摘要

A sound economic environment depends on an efficient financial system. For banks, the most important activity is credit risk management; banks use credit scoring to differentiate potential customers by risk level and provide different levels of monitoring and servicing. Many companies are looking at the inadequacies of past scoring models. They start from data sources, expecting to change the entire scoring ecosystem. Although, they want to take advantage of these opportunities and earn money. Under the change of the credit scoring environment, banks are gradually adjusting old methods of credit evaluation. This paper, from the perspective of the bank, considers the impact and development of community data on credit scores in the future. This study will not deal with individual ratings, but will focus on the entire customer base. The research will combine the variables that banks used to do credit scoring in the past with the personal friendship on social media data, to classify customers This study hopes to glimpse the current action plans of major banks from the experimental results and provide them with a feasible credit evaluation solution. The main research goals are as follows: (1) How much influence does the community have in daily life? Is it worth to being cited as a variable for credit scoring? (2) Exploring whether the credit classification covering social media data has a direct or indirect impact on credit assessment and risk control? Will this method reduce the amount of failures that caused by the model which only covered personal financial information? (3) Exploring whether a new credit scoring method with social media data can identify potential connections and profit from existing customers? (4) Exploring the new development and application level of the new credit scoring method including social media data. What kind of industries are suitable to be judging by this method, and which areas can be covered in the future? At the end of the study, the direction and application field that can be improved in the future are proposed, and it is expected to bring a more effective credit evaluation method to the bank.

並列關鍵字

Credit evaluation ； Credit score ； Social media ； Support vector machine ； Neural network ； Logistic regression

參考文獻

英文文獻

Google Scholar

【1】 Cios KJ, Pedrycz W and Swiniarski RW, “Data Mining Methods for Knowledge Discovery”, Boston, KluwerAcademic Publishers, 1998

Google Scholar

【2】 Caroline Haythornthwaite, “Strong, Weak, and Latent Ties and the Impact of New Media”, The Information Society, 18:385–401, 2002

Google Scholar

【3】 Chen-Guang Yang and Xiao-Bo Duan, “Credit risk assessment in commercial banks based on SVM using PCA”, International Conference on Machine Learning and Cybernetics, 2008

Google Scholar

【4】 D.J. Hand and W.E. Henley, “Statistical Classification Methods in Consumer Credit Scoring: a Review”, J. R. Statist. Soc. A (1997)160, 1997, pp.523-541

Google Scholar

國際替代計量

結合社群資料與現有信用評估參考指標進行信用分類

全文下載

主題瀏覽