行业智能>> Risk-Fraud-Analytics>> 返回
项目作者： diegoct6

项目描述 ：
Predictive modeling projects developed during the Risk & Fraud Analytics course (Master in Business Analytics & Big Data) at IE HST.
高级语言：
项目地址: git://github.com/diegoct6/Risk-Fraud-Analytics.git
创建时间： 2020-08-11T00:41:25Z
项目社区：https://github.com/diegoct6/Risk-Fraud-Analytics
开源协议：
下载


Risk & Fraud Analytics

Predictive modeling projects developed during the Risk & Fraud Analytics course (Master in Business Analytics & Big Data) at IE HST.

Akwnoledgements: I here make a great use of the starter code as well as other codes provided by Professor Manoel Gadi.

1. Competition for Online Fraud Dectection Model

This is the code that achived a maximum grade for the Kaggle style competition in which we had to develop a model to detect fraud for a startup. This startup operates in Peru with inititally 800 customers that later grew.

The approach I follow in my discovery for the best predictive model are roughly organized around the following steps:

  • Sampling and Performance Definition
  • Model Evaluation: KS and Gini
  • Models Trials: Logistic Regression, Naive Bayes, KNN, SVM, Random Forrest, XGBClassifier, LGBMClassifier, Neural Network, and CatBoostClassifier
  • Hyperparameter tunning
  • Feature Selection
  • Preprocessing the variables

I provide the code I used as well as a detailed logbook.