Global-Ecommerce-Retail-Trends-Analysis

Overview

This project involves analyzing global e-commerce trends and their impact on traditional retail. The analysis includes data preprocessing, outlier detection, Principal Component Analysis (PCA), Customer Lifetime Value (CLV) calculation, and a What-if analysis to simulate the effect of different pricing strategies.

Files Included

1. Datasets:

   • 1.csv, 2.csv, 3.csv: Raw e-commerce and retail datasets containing information about sales, customer transactions, and more.
   • tashi.csv: A combined and preprocessed dataset that merges 1.csv, 2.csv, and 3.csv. Download
   • pca_transformed_data.csv: Dataset after applying PCA for dimensionality reduction. Download

2. Reports:

   • Report.pdf: A comprehensive report detailing the analysis, visualizations, and insights.
   • Pre-Processing_Insights.pdf: A detailed document explaining the preprocessing steps taken, including outlier detection and handling missing values.

3. Notebook:

   • code.ipynb: Jupyter notebook containing Python code for data preprocessing, PCA, CLV calculation, and visualizations.

Project Steps

1. Data Preprocessing

• Data Integration: Datasets 1.csv, 2.csv, and 3.csv are combined into a single dataset tashi.csv.
• Handling Missing Values: Missing values in numerical columns were filled using the mean of each column.
• Outlier Detection and Removal: Outliers were detected using the Z-score method with a threshold of 2.5. These outliers were removed from the dataset.
• Normalization: Numerical columns were normalized using MinMaxScaler.
• Label Encoding: Categorical columns were label encoded using LabelEncoder.
• Feature Creation:
  • Total_Sales: Calculated by multiplying UnitPrice and Quantity.
  • Discount_Effectiveness: Ratio of discount_amount to Total_Sales.
  • Sales_per_Customer: Sum of total sales per customer.

2. Dimensionality Reduction with PCA

• Principal Component Analysis (PCA): Applied to the normalized dataset to reduce dimensionality while retaining 80% of the variance.

• Visualizations:

  • Scatter plot of the first two principal components.

  • Heatmap and boxplots to visualize PCA results.

3. Customer Lifetime Value (CLV)

• CLV Calculation: CLV was calculated for each customer based on average purchase value, purchase frequency, and retention rate.
• CLV Visualization:

  • Boxplots and violin plots were used to visualize CLV across different customer segments.

  • A heatmap of average CLV across customer segments was generated.

4. What-If Analysis

• Price Change Simulations: The effect of different price changes on CLV was simulated by modifying the UnitPrice variable. The results were visualized using line plots and histograms.
• Visualization of Impact: Line plots and heatmaps were created to illustrate the impact of different UnitPrice multipliers on CLV and total sales.

Running the Notebook

1. Open the Jupyter notebook code.ipynb in any Jupyter environment (e.g., JupyterLab, Google Colab).
2. Run the code cells in sequence to preprocess the data, apply PCA, calculate CLV, and generate the visualizations.

Dataset Files:

• Ensure that 1.csv, 2.csv, and 3.csv are available in the working directory.
• The notebook generates tashi.csv and pca_transformed_data.csv as outputs.

Key Insights

• Dimensionality Reduction: PCA effectively reduced the dataset dimensions while preserving most of the data's variance.
• CLV Segmentation: Customer Lifetime Value was calculated and segmented, highlighting differences in customer value across groups.
• Impact of Price Changes: The What-if analysis demonstrated how changes in UnitPrice affect CLV, providing insights into potential pricing strategies.

Visualizations

• Scatter Plot of Principal Components: Highlights clusters or separations in the dataset after applying PCA.
• Boxplot of CLV Segments: Showcases the distribution of CLV across different customer segments.
• Violin Plot of CLV Segments: Visualizes the density of CLV values across segments.
• Heatmaps: Used to display the relationship between CLV, total sales, and UnitPrice changes.
• Histograms: Demonstrate the frequency of CLV values for various price multipliers in the What-if analysis.

Contact

For any questions or suggestions, feel free to contact at [abbasitashfeen7@gmail.com]