pabelnl/mid-bootcamp-project

This project aimed to develop models that effectively predict customer acceptance of credit card marketing offers.

Below is a breakdown of the steps employed:

Data Preparation

1. Data Import and Shape Verification: Imported spreadsheet data for analysis. Checked the dimensions to get an initial feel for the size of the dataset.

2. Identifying the Target Variable: Based on the requirements, determined 'offer_accepted' as the target that the model must predict.

3. Formatting and Exploratory Analysis: Renamed columns with suitable formatting. Then, analyzed data types and identified null values in features.

4. Converting to Categorical Features: Variables like the number of bank accounts, credit cards, household size, etc., were identified as potential categorical features.

5. Removing Unnecessary Features: Dropped non-essential features (e.g., customer number).

6. Handling Missing Values: Employed a random sample imputer strategy to replace missing values.

7. Outlier Treatment: Created a function to identify and adjust outliers, focusing on continuous numerical features.

Data Transformation and Feature Engineering

1. Correlation Analysis: Examined how variables relate to each other using a correlation heatmap. Removed a feature to address high colinearity.

2. Chi-Squared Test: Employed this test to assess relationships between categorical features and the target.

Encoding:

• Categorical features underwent transformation by one-hot encoding

• Applied standard scaling to continuous numerical features.

• Data Splitting: Divided the data into training and testing sets to allow for model development and evaluation.

Modeling and Evaluation

Baseline Model: Started with a Logistic Regression model to establish an initial reference for performance.

Model Comparison: Trained and evaluated additional models with Decision Trees and K-Nearest Neighbors algorithms.

KNN Optimization: Investigated model performance variations based on a range of 'K' values (neighbors) in KNN to fine-tune performance.