Walmart-Weekly-Sales-ML-Algorithm

The solution to One of Kaggle's Datasets: https://www.kaggle.com/datasets/yasserh/walmart-dataset
An interactive Jupyter Cell for finding Weekly Sales based on input!
Accuracy: 92.677%

Key Components of Solution

1. Data Preprocessing:
   1. Made numpy array of features from Holiday_flag to Unemployment. Shape: (m, 5)
   2. Store is a categorical data and cannot be quantized as an ordinal number and hence used one-hot encoding to form a 45 column ndarray for each shop, 1 representing presence and 0 for absence. Shape: (m, 45)
   3. Made another ndarray which extracted month, year as dates from data
   4. Concatenated important features, dates, one hot encoded shops and stored as X_train.
   5. Seperate ndarray for storing output of each examples and stored as y_train. Reshaped it into 2d array for matrix multiplications.
2. Training:
   1. Standardized using z score: X_train and y_train for feature scaling
   2. Mean Squared Error
   3. Gradient Descent
   4. r2 score for accuracy metrics
   5. Interactive cell for predicting new weekly sales based on respective features.
3. Used numpy matrix multiplications for faster computations and cleaner code.