# Health-Insurance-Cross-Sell-EDA-and-Machine-Learning-Modeling
This is a note book of exploratory data analysis on cross selling of health insurance customers on vehicle insurance product and using machine learning to predict whether a customer is interested or not in vehicle insurancen
## Background Information :
- an Insurance company that provide Health Insurance to its customers, usually they offer other insurance product to the customers through diffirent kind of marketing channel. In this case we will build a model to predict whether the policyholders (customers) from past year will also be interested in Vehicle Insurance provided by the company.
## Problem Statement:
- Un optimize customer reachout process, many insurance worker spend a lot of their time having meeting with prospective client without knowing the probabily of that customer to buy the insurance product
## Business Goals:
- Building a model to predict whether a customer would be interested in Vehicle Insurance is extremely helpful for the company because it can then accordingly plan its communication strategy to reach out to those customers and optimise its business model and revenue
## Business Question:
- How does age of a vehicle determing the response of vehicle insurance advertisment
- How to attract customers from different generation
- what's the major factor that make a health insurance customer not intersted with vehicle insurance
- What's the best machine Learning modeling for this Cross Sell case
## Workflow :
- Data Cleaning
- Recategorize Data
- Binning
- Exploratory Data Analysis to Answer business Question
- Feature Engineering & Selection For Machine Learning Process
- Encoding all the categorical features
- Checking correlation between dependent and independet variable
- Feature Selection
- Model Building :
- Splitting data into Training and Testing
- Apllying the **SMOTE** (Synthetic Minority Oversampling Technique) to the minority target since the data is imbalance
- Creating base model of classification algorithm ( Logistic Regression, KNN Classifier, Decision Tree Classifier, Random Forest Clasifier)
- Check The Evaluation matrix for all the base model
- HyperParameter tuning
- Checking Evaluation Matrix for tuned Model
- Choose which model has the best recall score for this case
<details>
<summary><h2> Click to see the conclusion and recommendation </h2></summary>
## Conclusion:
- From this dataset of health insurance customers almost **95%** of customers have a vehicle age that's less than 2 years. from our analysis, customers who has more than 2 years of vehicle age are more interested with vehicle insurance advertisment, while customers who has less then one year of vehicle age, only **4%** of them are actually interesred with vehicle insurance
- We found out that customer who already have vehicle insurance are almost have no interest in another vehicle insurance. Our analysis shows that **99.9% of customers that have a vehicle insurance is not interested in another vehicle insurance**, while customer who doesn't have a vehicle insurance **22.5 %** of them are interested with vehicle insurance
- we also found out that a newer vehicle are more likely to have a vehicle insurance, with vehicle that's less than one year **66% of those are insured** , vehicle that's older than one year but less than 2 years are **33% insured**, while less than **one percent** of vehicle that's older than 2 years are insured. This should explain why customer who owns a newer vehicle are less likely to be intersted with insurance promotion, because they probably alredy have one.
- **Customers who never had vehicle damaged** only 0.5 % of those customers are intersted with vehicle insurance, **87%** of customers who never had any vehicle damaged already have a vehicle insurance
- **Which Customer Generation are less likely to be intersted with vehicle insurance** the answer is **Millenials** (people in age group of 18 - 34) only **6% of millenials are actually interested with vehicle insurance**, and why is so?
1. almost **63% of millenials already have vehicle insurance**, from our analysis before owning vehicle insurance is a major factor why someone is not interested with another vehicle insurance
2. **90% of millenials have a vehicle that's less than one year of age**, and from our analysis before that vehicle that's less than one year are **66% already insured**
This conculed that millenials are more likely to already have a vehicle insurance before our vehicle insurance team approached, and that's a major factor why millenials are least likely to be interested with our vehicle insurance, because they already have one
#### So who's actually interested with our vehicle insurance ?
From the responses there are **12 % of our health insurance customers are interested with the vehicle insurance product** but who are those people?
1. **First, Customer who does not have a a vehicle insurance**, out of all customers who does not have a vehicle insurance **22.5 %** of them says that they're interested with vehicle insurance product
2. **customers who has vehicle that's older than 2 years** our analysis before mentioned that only less than **one percent** of car that's older than 2 years are previouly insured, by not having a vehicle insurance they're more likely to be intersted with our vehicle insurance, our data show's that customer who has car that's more than 2 years are **7 times** more likely to be intersted with vehicle insurance compared to customer who own a vehicle less than one year
3. **customers who have had a vehicle damaged in the past** from our analysis we found out that **97%** customers who actually intersted with vehicle insurance have had their vehicle damaged in the past
1. **95 % of customers who have had vehicle damage in the past still doesn't have a vehicle insurance**
**Which Customer Generation that's most likely to be interested in Vehicle insurance ?**
- **GEN X** (Age Gen X : 35 - 50)
- our analysis shows that GEN X has the highest percentage to be intersteed with vehicle insurance, to be precise, **21 %** of GEN X are interested with vehicle insurance
- This might be because **72% GEN X** does not have a vehicle insurace, and GEN X has the highest percentage of vehicle damager the past **(67%)** among other generation
<b>Machine learning could predict whether a customer would be interested or not towards vehicle insurance product with recall 0.965 out of 1</b>
Using logiistic regression that has been tuned, we focus more on recall instead of accuracy here because of we want to reduce the false negative ( The customer who actually interested but the model predicted that customer is not intrested). Since this kind of problem could lead into lost of potential revenue
## Recommendation
### 1. Work with dealership to capture millenial market
as we know from the analysis that millenials are less likely to be intersted with vehicle insurance because of most of them have a vehicle that's less than one year of age, and vehicle with less then one year of age are most likely to be insured so in conclution they already have one, and so they're not interested. By working together with a dealership that sells a brand new car, we could tackle this problem, our insurance company could have a bundling product of brand new vehicle and a free promotional vehicle insurance for certain period of months. we hope that by working together with vehicle dealership we could target more millenials customers.
### 2. Target & Educate Customers Who had Vehicle Damage in the past
**95%** customers who have had a vehicle damaged in the past still does not have a vehicle insurance this is a gold mine for our vehicle insurance, since customers are more likely to be interested in vehicle insurance if they've a vehicle damage in th
没有合适的资源?快使用搜索试试~ 我知道了~
健康保险交叉销售 EDA 和机器学习建模
共4个文件
ipynb:2个
md:1个
csv:1个
4 下载量 81 浏览量
2023-03-04
15:52:58
上传
评论
收藏 4.93MB ZIP 举报
温馨提示
健康保险交叉销售 EDA 和机器学习建模 这是一本关于健康保险客户交叉销售车辆保险产品并使用机器学习预测客户是否对车辆保险感兴趣的探索性数据分析笔记本 背景资料: 向客户提供健康保险的保险公司,通常他们通过不同类型的营销渠道向客户提供其他保险产品。在这种情况下,我们将建立一个模型来预测过去一年的保单持有人(客户)是否也会对公司提供的车辆保险感兴趣。 问题陈述: 未优化客户联系流程,许多保险工作者花费大量时间与潜在客户会面,却不知道该客户购买保险产品的可能性 业务目标: 建立一个模型来预测客户是否会对车辆保险感兴趣,这对公司非常有帮助,因为它可以相应地规划其沟通策略以接触这些客户并优化其业务模型和收入 业务问题: 车龄如何决定车险广告反应 如何吸引不同世代的顾客 使健康保险客户对车辆保险不感兴趣的主要因素是什么 这个交叉销售案例的最佳机器学习建模是什么 工作流程: 数据清洗 重新分类数据 分级 探索性数据分析以回答业务问题 机器学习过程的特征工程和选择 编码所有分类特征 检查因变量和独立变量之间的相关性 特征选择 建筑模型 : 将数据拆分为训练和测试 由于数据不平衡,将SMOT
资源推荐
资源详情
资源评论
收起资源包目录
Health-Insurance-Cross-Sell-EDA-and-Machine-Learning-Modeling-main.zip (4个子文件)
Health-Insurance-Cross-Sell-EDA-and-Machine-Learning-Modeling-main
Health insurance cross sell prediction Machine Learning Modeling Modeling.ipynb 389KB
train.csv 20.44MB
Health Insurance Cross Sell EDA.ipynb 559KB
README.md 14KB
共 4 条
- 1
资源评论
小夕Coding
- 粉丝: 5844
- 资源: 461
上传资源 快速赚钱
- 我的内容管理 展开
- 我的资源 快来上传第一个资源
- 我的收益 登录查看自己的收益
- 我的积分 登录查看自己的积分
- 我的C币 登录后查看C币余额
- 我的收藏
- 我的下载
- 下载帮助
安全验证
文档复制为VIP权益,开通VIP直接复制
信息提交成功