xx-craziness-xx/shipment-price-prediction

SHIPMENT PRICE PREDICTION

The market for supply chain analytics is expected to develop at a CAGR of 17.3% from 2019 to 2024,more than doubling in size

The main goal is to predict the supply chain shipment pricing

Screenshots

Dataset

https://www.kaggle.com/divyeshardeshana/supply-chain-shipment-pricing-data

Installation

• Create a new environment

• Copy the downloaded files or clone the files from github to your environment

• pip3 install -r requirements.txt

• python app.py

Data Preprocessing and Feature Engineering

• Some values of column Weight and Freight price were mapped to a specific Id in the dataset like See DN-4307 (ID#:83920) so using some custom function retrived the values from the column

• The null value was filled by using median value

• Outliers were removed by using IQR

• Created a new feature distance ie., distance from the manufacturing country to destination country, for that used Geopy to get the latitude and longitude of different countries to get the distance

• Few features were dropped using correlation

Model Creation and Model Selection

• The categorical values were encoded using Category Encoders library

• Used Linear Regression as a base model score

• For advanced model selection and hyperparameter tuning I used FLAML AutoML and chose ExtraTreesRegressor as the best model with a R2 score of 85.01% and RMSE of 3391.33 on the test dataset

Deployment

Deployed in Heroku

https://shipment-price-prediction.herokuapp.com/