Project Overview:

This project is focused on identifying the best Halloween candies in terms of both popularity and affordability. Using data science techniques such as Principal Component Analysis (PCA) and K-Means clustering, I explore the various candy attributes (e.g., chocolate, fruity, nutty, bar form) to classify candies into meaningful groups. Additionally, a custom 'popular_vs_cheap' score is calculated to highlight the candies that are not only well-liked but also cost-effective. The final output is presented in a PowerPoint-style visualisation that helps inform candy selection for Halloween.

Key Steps:

1. Data Preparation:

   • Cleaning and preprocessing the candy dataset (removing unnecessary columns, renaming for clarity).
   • Creation of a custom 'popular_vs_cheap' metric based on candy price and win percentage.
   • Creation of a 'price_range' column to categorise candies into different price tiers based on the price percentile: - 'Very Low' for the lowest 25% of prices, - 'Low' for the next 25%, - 'Medium' for the middle 25%, - 'High' for the highest 25%.

2. Dimensionality Reduction:

   • Applied Principal Component Analysis (PCA) to reduce the feature space and uncover the most important dimensions in the data (e.g., fruity vs. chocolate, bar vs. bag).

3. Clustering:

   • K-Means clustering was used to group candies into three clusters:

     1. Chocolate Nutty Sweets
     2. Fruity Bags
     3. Chocolate Flavoured Bars

4. Visualisation:

   • A combination of Seaborn and Matplotlib was used to generate scatter plots and heatmaps.
   • Jitter was added to scatter plots to avoid overlapping points for better clarity.
   • Results were visualised in a fun and engaging manner, including the presentation of findings in a Halloween-themed dashboard for easy interpretation.

Tools and Libraries:

• Pandas: For data manipulation, cleaning, and preparation.
• NumPy: Used for numerical operations and adding jitter to the PCA components.
• Scikit-learn: For applying PCA and K-Means clustering.
• Matplotlib and Seaborn: For data visualisation, including heatmaps and scatter plots.
• PowerPoint-style Visualisation: The final results are presented in a PowerPoint-like format with thematic Halloween elements.

Final Deliverable:

The project concludes with a Halloween-themed visualisation summarising the best candy choices, balancing both popularity and affordability. The visualisation is suitable for decision-making in candy selection, helping keep costs down while still being a hit with the kids!