Air Pollution Prediction: Time-series clustering & ARIMA

Link to Data

https://drive.google.com/drive/folders/1SnDNT0aDZ_vn4ily36M24nbel2Y-U1-8?usp=sharing

How to run | Clone the whole repository

1. The code files are ipnb (IPython notebook) verion--Python3
2. You can open the project in any IDE that supports .ipnb Eg. Jupyter/Anaconda Download here. You may need to download some libraries that are used in the programs. If you are using Jupyter/Anaconda, you can install the libraries - conda install library_name
3. The main programs for Month-wise clustering & state-wise clustering can be found in 'main' folder.
4. Download the data from here.
5. Change the data folder location path as per your system. The change has to be made in 1st or 2nd cell of MonthWiseClustering.ipynb and StateWiseClustering.ipynb
6. Point 5 applies to other files also, where data folder is being accessed.

Repository Structure Description | Folder Name - Description

1. Evaluation - Includes programs developed for evaluation of different algorithms using libraries, mainly scikit-learn. LSTM code for evaluating the behavior of our data is taken from https://machinelearningmastery.com.
2. Reports - Contains the intermediate reports, poster and final report.
3. Main - Contains four files.
   1. CreateTimeSeries - Pre-process the data and create time-series data ready to be fed into custering algorithms
   2. Euclidean vs DTW - Evaluating Euclidean distance and Dynamic Time Warping for two time-series'.
   3. MonthWiseClustering - Implementing K-Means with DTW on month-wise time-series data.
   4. StateWiseClustering - Implementing K-Means with DTW on state-wise time-series data.
4. Preprocessing - Pre-processing raw data
5. state Data - Data produced by pre-processing, contains state-wise csv files.