ChurnWise

The churn prediction app leverages Snowflake and Streamlit, utilizing advanced machine learning algorithms to analyze customer behavior and identify patterns that may indicate potential churn.

The Streamlit-based interface ensures user-friendly interactions, catering to both data professionals and business stakeholders. This app empowers organizations to make informed decisions and optimize their customer retention efforts.

Features included:

1. Churn Propensity:
Compute churn probability score for all the customers under selection.

2. Churn Classification:
Classify churn probability score into High, Medium and Low.

3. Feature Weights:
Provide weights and directions for model attributes.

4. Model Explanability:
Provide model Explanability to understand the computation of churn probability at each customer level.

5. Multi-Business Sector:
The Customer Churn Propensity model may be used for mutiple business sectors (Banking, Insurance, Financial Services etc.).

Expected workflow for client specific outputs:

* The client does not have to handle any development activities as it is entirely handled by the NSEIT team.

* Consumer data are kept private and secure. After the app is installed, it is recommended by the provider to grant the following privileges as needed.

Steps to Use

1. Variable/Feature Description:

* CreditScore: Represents the credit score of the customer.
* Geography: Indicates the geographical location of the customer.
* Gender: Denotes the gender of the customer.
* Age: Represents the age of the customer.
* Tenure: Denotes the number of years the customer has been associated
with the institution.
* Balance: Represents the account balance of the customer.
* NumOfProducts: Indicates the number of products (e.g., banking products)
the customer has with the bank.
* HasCrCard: Indicates whether the customer has a credit card (1 for Yes, 0
for No).
* IsActiveMember: Denotes whether the customer is an active member (1 for
Yes, 0 for No).
* EstimatedSalary: Represents the estimated salary of the customer.

2. It is mandatory to keep same column names to run this app successfully.

3. Customer can initiate the process by installing the churn prediction application directly from the Snowflake Marketplace, ensuring a seamless integration into their existing Snowflake environment.

4. Customer can then grant select and insert access to the specific tables/views containing scoring data and final results with columns matching the following definitions:

* scoring_data:

* customer_id (varchar(100))
* creditscore (number(10,0))
* geography (varchar(1000))
* gender (varchar(100))
* age (number(5,0))
* tenure (number(5,0))
* balance (number(28,10))
* numofproducts (number(10,0))
* hascrcard (number(1,0))
* isactivemember (number(1,0))
* estimatedsalary (number(28,10))

* final_result:

* customer_id (varchar(100))
* creditscore (number(10,0))
* geography (varchar(1000))
* gender (varchar(100))
* age (number(5,0))
* tenure (number(5,0))
* balance (number(28,10))
* numofproducts (number(10,0))
* hascrcard (number(1,0))
* isactivemember (number(1,0))
* estimatedsalary (number(28,10))
* probability (number(28,10))
* churn_category (varchar(20))

5. Customer can start using the application for customer churn prediction.

本客户流失预测应用依托Snowflake与Streamlit，借助先进机器学习算法 (Machine Learning Algorithms) 分析客户行为，识别可能预示潜在客户流失的模式。

基于Streamlit的交互界面可提供友好的操作体验，同时适配数据专业人士与业务利益相关者的使用需求。本应用可助力企业制定科学决策，优化客户留存工作。

核心功能如下：
1. 客户流失倾向 (Churn Propensity)：为选定范围内的所有客户计算流失概率得分。
2. 客户流失分级 (Churn Classification)：将流失概率得分划分为高、中、低三个等级。
3. 特征权重 (Feature Weights)：为模型属性提供权重与作用方向。
4. 模型可解释性 (Model Explanability)：提供模型可解释性功能，便于理解单客户维度下的流失概率计算逻辑。
5. 多业务场景适配 (Multi-Business Sector)：本客户流失倾向模型可适配多个业务领域（如银行、保险、金融服务等）。

面向客户的定制化输出工作流程如下：
* 客户无需开展任何开发工作，所有开发事宜均由NSEIT团队全权负责。
* 客户数据将得到严格的隐私保护与安全保障。应用部署完成后，服务提供商建议根据实际需求授予以下权限。

使用步骤：
1. 变量/特征说明 (Variable/Feature Description)：
* 信用评分 (CreditScore)：代表客户的信用得分。
* 所在地区 (Geography)：指明客户的地理位置。
* 客户性别 (Gender)：标识客户的性别。
* 客户年龄 (Age)：代表客户的年龄。
* 合作时长 (Tenure)：指客户与机构的合作年限。
* 账户余额 (Balance)：代表客户的账户余额。
* 持有产品数量 (NumOfProducts)：指客户在银行持有的产品（如银行产品）数量。
* 是否持有信用卡 (HasCrCard)：标识客户是否持有信用卡（1代表是，0代表否）。
* 是否为活跃会员 (IsActiveMember)：标识客户是否为活跃会员（1代表是，0代表否）。
* 预估薪资 (EstimatedSalary)：代表客户的预估薪资。
2. 为确保应用正常运行，必须使用完全一致的列名。
3. 客户可直接从Snowflake Marketplace下载安装本客户流失预测应用，即可实现与现有Snowflake环境的无缝集成。
4. 随后客户可向包含评分数据与最终结果的特定表/视图授予查询（SELECT）与插入（INSERT）权限，这些表/视图的列需符合以下定义：
* 评分数据 (scoring_data)：
* 客户ID (customer_id)：varchar(100)类型
* 信用评分 (creditscore)：number(10,0)类型
* 所在地区 (geography)：varchar(1000)类型
* 客户性别 (gender)：varchar(100)类型
* 客户年龄 (age)：number(5,0)类型
* 合作时长 (tenure)：number(5,0)类型
* 账户余额 (balance)：number(28,10)类型
* 持有产品数量 (numofproducts)：number(10,0)类型
* 是否持有信用卡 (hascrcard)：number(1,0)类型
* 是否为活跃会员 (isactivemember)：number(1,0)类型
* 预估薪资 (estimatedsalary)：number(28,10)类型
* 最终结果 (final_result)：
* 客户ID (customer_id)：varchar(100)类型
* 信用评分 (creditscore)：number(10,0)类型
* 所在地区 (geography)：varchar(1000)类型
* 客户性别 (gender)：varchar(100)类型
* 客户年龄 (age)：number(5,0)类型
* 合作时长 (tenure)：number(5,0)类型
* 账户余额 (balance)：number(28,10)类型
* 持有产品数量 (numofproducts)：number(10,0)类型
* 是否持有信用卡 (hascrcard)：number(1,0)类型
* 是否为活跃会员 (isactivemember)：number(1,0)类型
* 预估薪资 (estimatedsalary)：number(28,10)类型
* 流失概率 (probability)：number(28,10)类型
* 流失等级 (churn_category)：varchar(20)类型
5. 至此，客户即可开始使用本应用开展客户流失预测工作。

搜集汇总

背景与挑战

背景概述

ChurnWise是一个客户流失预测数据集，利用机器学习算法分析客户行为并预测流失概率，适用于多个业务领域。它提供流失评分、分类、特征权重和模型解释功能，帮助组织优化客户保留策略。