This project is set out to create a model to forecast the likely event ticket sales prior to an event based on previous sales data to start date of an event to enable users decide if more promotion to reach a targeted booking sales is required. This repository consists of mainly Jupyter Notebook files.
| Name | Brief |
|---|---|
| Streamlit | An open source app framework for building beautiful data and ML apps |
| Numpy | Python library for numerical computation |
| Pandas | Python library for data manipulation and analysis |
| JSON | Python library using JSON API client library |
| datetime | Python library using new date and time classes |
| XGBoost | Python library using XGBoost Python Package |
| time | Python library using time module |
| itertools | Python library itertools module |
| Scipy | An open-source software for mathematics, science, and engineering |
| Matplotlib | Python library for creating static, animated, and interactive visualizations |
| Seabron | Python library based on matplotlib |
| Scikit-learn | Python library for machine learning |
To get started, clone the Git repo onto your laptop using this command on your terminal in your preferred directory:
git clone https://github.com/ut6817150/Forecasting_Web_App.git
| File | Description |
|---|---|
| /Sales-Forcast-App | Root directory |
| /01_import/01_import_data.ipynb | This notebook merges the 2 data |
| /02_Preprocessing/01_Sampling.ipynb | This script creates a dataset of the sampled dataset and applies some initial cleaning steps |
| /02_Preprocessing/02_Data_audit.ipynb | This script takes an audit on the dataset. It looks at how many missing values each feature has, how many unique values the categorical features have then applies pre-processing steps to remove outliers and drop features with too many missing values |
| /02_Preprocessing/03_corr_and_chi2.ipynb | This script finds the correlations between numerical features of the dataset and performs a chi-squared test on all dataset variables |
| /03_EDA/01_EDA.ipynb | This script performs exploratory data analysis on the dataset |
| /04_Baseline_Model/01_Baseline_Model.ipynb | This script applies a linear regression model to get weekly sales distribution, giving a baseline performance for comparison |
| /05_Modeling/01_Modeling.ipynb | 2 different supervised learning regression models are assessed to understand how well they could predict weekly sales distribution |
| /06_Web_Application /Dashboard.py | Creates the dashboard page for the streamlit framework web application |
| /06_Web_Application /Prediction.py | Creates the prediction page for the streamlit framework web application |
| /config-example.json | JSON file containing elements for loading files |
| /config-explanation.md | Explains what each element of the config JSON file refers to |
Read config-example.json and config-explanation.md to understand the loading of files template.
To install the python dependencies, create a Python virtual environment for the model:
python3 -m venv ~/SalesApp
Run this command whenever you open a new Terminal window/tab to activate the environment.
# Unix/MacOs
source SalesApp/bin/activate
# Windows
.\SalesApp\Scripts\activate
Next, to install the python dependencies run the following:
# Install Python dependencies
python3 -m pip install pip --upgrade
python3 -m pip install -r requirements.txt
Next, run each file from 01_import to 05_Modeling.
Below demonstrates how the web application is used to obtain sales predictions.
-
User inputs their event type, event start date, booking period (Weeks to event start date) and clicks the predict button.
-
The formatted inputs are sent to streamlit API and returns all corresponding sales prediction, cumulated sales prediction, cumulated sales percentage, and the weeks to event as a table with the plotted predicted sales Vs weeks to event.
-
The formatted user inputs are sent to streamlit API and also returns summary of output from the table and steps to take with evaluating performance of event towards the start.
