Project Title: Shipping Forecasting for Shoppedia e-commerce.
Project Goal: to increase Delivery on Time rate from 40% to 60%.
Project Objective: to predict whether the packages from Shoppedia e-Commerce company arrived on time or not based on the historical dataset.
Dataset: ID, Warehouse Block, Mode of Shipment, Customer care calls, Customer ratings, Cost of the Product, Prior purchases, Product importances, Gender, Discount offered, and Weight.
Insight summary:
- High Discounts correlate positively with Late Deliveries.
- Weight and Price of Products impact Timely Deliveries.
- Warehouse Blocks and Mode of Shipment Affect Discounts and On-Time Deliveries.
Data Preprocessing Summary:
- Tidak terdapat missing value, duplicate data, dan invalid value.
- Dilakukan feature extraction dengan nama Weight_class dari feature berat barang (Weight_in_gms), menjadi 4 kategori berdasarkan beratnya.
- Dilakukan feature selection berdasarkan multicolinearity, variance dan entropy, dan Information Value. Feature yang masih digunakan yaitu Customer_care_calls, Cost_of_the_Product, Discount_offered, Prior_purchases, Product_importance, dan Weight_class.
- Dilakukan train/test data split untuk persiapan machine learning modeling.
- Tidak dilakukan handling outlier karena dataset yang kecil, transformasi data menggunakan metode yang robust terhadap outlier.
- Transformasi dan scaling data dilakukan pada feature Cost_of_the_Product dan Discount_offered.
- Dilakukan Label Encoding untuk Feature Encoding untuk feature kategorikal (Product_importance).
- Tidak dilakukan handling class imbalance, karena jumlah nilai pada dataset tidak masuk dalam kriteria class imbalance.