本文關鍵詞：P2P網貸平臺的信用風險評級研究　出處：《北方工業(yè)大學》2017年碩士論文　論文類型：學位論文

  更多相關文章： 信用風險 網貸平臺 風險評級 機器學習

【摘要】：作為互聯網金融的重要組成部分,P2P網絡借貸在我國得到快速的發(fā)展,網貸平臺數量不斷增加,但是在監(jiān)管缺失的環(huán)境下,問題平臺層出不窮,給整個網貸行業(yè)發(fā)展和社會穩(wěn)定帶來了巨大的負面影響。本文在信用風險評級理論基礎上,借鑒企業(yè)和銀行的信用風險評級方法,結合網貸平臺的特點,選擇Credit Metrics模型作為研究方法。在不同時間點分別對網貸平臺做信用評級,建立信用風險轉移矩陣。對網貸平臺做信用風險評級時,首先分析了網貸平臺的運營模式和盈利模式,結合評級機構的指標體系,選擇反映網貸平臺的信用風險的指標。其次分別使用非監(jiān)督學習中的聚類分析方法和有監(jiān)督學習中的決策樹、SVM和提升算法對網貸平臺的信用風險進行衡量。在未知網貸平臺是否違約時,使用聚類分析方法對網貸平臺做分類,根據指標選擇確定各類別的信用等級,完成網貸平臺的信用風險評級,并對結果進行跟蹤,結果表明:研究的網貸平臺樣本中,出現違約的平臺占7%左右,且都是評級結果較低的平臺,借貸利率與評級結果呈負相關。在獲得問題平臺樣本后,使用決策樹、SVM和提升算法對網貸平臺的信用風險做評級,結果顯示:有監(jiān)督學習算法優(yōu)于聚類分析,同時該類算法可實現對網貸平臺信用風險的預測,其中提升算法的準確率高。使用主成分分析對同樣的網貸平臺樣本做評級,準確率沒有聚類分析高。使用Adaboost算法對網貸平臺借款人的信用風險做研究,完善網貸平臺的信用風險評級。最后使用提升算法得到的信用風險評級結果,建立網貸平臺的信用風險轉移矩陣。加入回收率,衡量網貸平臺違約后的償還能力。信用風險轉移矩陣表征網貸平臺的違約概率,結合回收率,可預估網貸平臺的信用風險大小。研究結果顯示:利率的大小和評級結果呈反向關系,借款利率越低,網貸平臺的信用等級越高,相反則越低。對網貸平臺的信用風險的評級,可預估網貸平臺違約的風險大小和整個行業(yè)的潛在風險,使監(jiān)管部門更好的管理網貸平臺,防范網貸行業(yè)風險的發(fā)生,同時可為投資者提供決策參考。
[Abstract]:As an important part of Internet finance, P2P network lending has been developing rapidly in China. The number of network lending platforms is increasing, but in the environment of lack of supervision, problem platforms emerge in endlessly. On the basis of the credit risk rating theory, this paper draws lessons from the credit risk rating methods of enterprises and banks, combined with the characteristics of the network lending platform. The Credit Metrics model is chosen as the research method. The credit rating of the network loan platform is done at different time points, and the credit risk transfer matrix is established. When the credit risk rating of the network loan platform is made, the credit risk rating of the network loan platform is made. First of all, it analyzes the operation model and profit model of the network loan platform, combined with the index system of the rating agencies. Select the index which reflects the credit risk of the network loan platform. Secondly, use the clustering analysis method in the unsupervised learning and the decision tree in the supervised learning. SVM and upgrade algorithm to measure the credit risk of the network loan platform. In the unknown network loan platform default, the use of clustering analysis to classify the network loan platform, according to the selection of indicators to determine the credit rating of each category. Complete the credit risk rating of the network loan platform, and track the results, the results show that: in the sample of the network loan platform, the default of the platform accounted for about 7%, and are the lower rating platform. The loan interest rate is negatively correlated with the rating results. After obtaining the sample of the problem platform, the credit risk of the network loan platform is rated using decision tree SVM and upgrade algorithm. The results show that the supervised learning algorithm is better than the clustering analysis, and this kind of algorithm can predict the credit risk of the network loan platform. Among them, the accuracy of the improved algorithm is high. Using principal component analysis (PCA), the sample of the same network loan platform is rated. The accuracy of clustering analysis is not high. Adaboost algorithm is used to study the credit risk of loan platform borrowers. Finally, using the credit risk rating results obtained by the upgrade algorithm, the credit risk transfer matrix of the network loan platform is established, and the recovery rate is added. The credit risk transfer matrix represents the default probability of the net loan platform, combined with the recovery rate. The research results show that the size of interest rate and rating results show a reverse relationship, the lower the borrowing rate, the higher the credit rating of the network lending platform. On the other hand, the lower the credit risk rating of the network loan platform, the risk of default and the potential risk of the whole industry can be estimated, so that the regulatory authorities can better manage the network loan platform. To prevent the occurrence of network loan industry risks, and to provide investors with decision-making reference.
【學位授予單位】：北方工業(yè)大學
【學位級別】：碩士
【學位授予年份】：2017
【分類號】：F832.4;F724.6

【參考文獻】

相關期刊論文 前10條

1 徐榮貞;殷元星;王帥;;P2P網絡借貸平臺運營模式及風險控制思考——基于信息不對稱視角[J];財會月刊;2017年05期

2 李琪;王振;;P2P網貸平臺風險和標的風險的辨析[J];上海金融;2016年10期

3 嚴復雷;李浩然;;P2P網貸平臺信用風險影響因素分析[J];西南金融;2016年10期

4 李戰(zhàn)江;句芳;修長柏;喬光華;;銀行信用風險小樣本評級模型的構建[J];統(tǒng)計與決策;2016年09期

5 王丹;張洪潮;;P2P網貸平臺信用風險評級模型構建[J];財會月刊;2016年09期

6 于曉虹;樓文高;;基于隨機森林的P2P網貸信用風險評價、預警與實證研究[J];金融理論與實踐;2016年02期

7 王珂;孟海麗;楊全;;模糊Credit Metrics模型及其在信用風險評估中的應用[J];金融理論與實踐;2016年02期

8 楊秀云;蔣園園;段珍珍;;KMV模型在我國商業(yè)銀行信用風險管理中的適用性分析及實證檢驗[J];財經理論與實踐;2016年01期

9 李琦;曹國華;;基于Credit Risk+模型的互聯網金融信用風險估計[J];統(tǒng)計與決策;2015年19期

10 李敏芳;田晨君;;基于因子分析法的我國P2P網貸平臺評級研究[J];湖北經濟學院學報(人文社會科學版);2015年06期

，

本文編號：1366580