The Weather is a significant factor influencing product sales, especially in retail stores. This project focuses on understanding the impact of various weather conditions on burger sales in a retail chain with stores across ten regions. The primary objective is to forecast burger sales based on historical data. The project includes deploying the model using AWS Sagemaker.
- Fetch data from SQL server hosted on Amazon RDS to Python.
- Analyze the influence of weather on burger sales.
- Build LSTM model on AWS Sagemaker.
- Forecast next day's sales based on historical data.
- Deploy the model using AWS Sagemaker.
- Data period: 1/1/2014 to 15/9/2020.
- Total rows: 24,500, Columns: 8.
- Data stored in MySQL server hosted on Amazon RDS.
- Language:
Python - Libraries:
pandas,numpy,tensorflow,boto3,pymysql,matplotlib,sklearn,tarfile,keras,lightgbm - Platform:
AWS Sagemaker,AWS S3
- Loading Data:
- Using pymysql library to fetch data from the MySQL server.
- Exploratory Data Analysis:
- Plotting date vs sales graph based on regions.
- Autocorrelation and partial correlation plot.
- Plotting average, maximum, and minimum of features.
- Scatterplot and histograms of the features.
- Data Preprocessing:
- Windowing the Data:
- Train-Validation Split:
- Data Rescaling using QuantileTransformer:
- Building Baseline Model (lightgbm) and Feature Importance:
- Building LSTM Model:
- Predictions for One Day into Future:
- Creating an Endpoint and Deploying the Model:
Burger_data.csv: Data used in the project.burger_stores.ipynb: Main file for training and deploying the model using Sagemaker.Config.yml: Store DB connection credentials.empty_train.py: Empty file created during model deployment on Sagemaker.Predict_from_endpoint.ipynb: File for predictions using the endpoint.Requirements.txt: File with required libraries and versions.utils.py: File containing functions called inburger_stores.ipynb.