Forecasting temperature and rainfall using deep learning for the challenging climates of Northern India

1. Description
This project presents a deep learning-based approach to forecast temperature and rainfall for the challenging climates of Northern India, focusing on weather data from Jammu, Srinagar, and Ladakh regions. The model utilizes Recurrent Neural Networks (RNN) and Long Short-Term Memory (LSTM) architectures to improve the accuracy of predictions over traditional statistical methods.

2. Dataset Information
Source: Indian Meteorological Department (IMD) Pune, accessed via the Data Supply Service URL:https://dsp.imdpune.gov.in/data_supply_service.php 
Time Period: January 1, 2010, to December 31, 2023.
Stations: Jammu, Srinagar, and Ladakh.
Variables:
Rainfall (mm)
Minimum Temperature (°C)
Maximum Temperature (°C)
Data Points: 8,401 time-series records per station.

3. Code Information
The code is organized into the following sections:
Preprocessing: Data cleaning, normalization, and train-test splitting.
Model Training: Implementation of RNN and LSTM models with hyperparameter tuning.
Evaluation: Model performance assessment using Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE) and Mean Absolute Percentage Error (MAPE).

4. Usage Instructions
To execute the code follow the below mentioned instructions:
a. Run weather.ipynb ,weathertr.ipnb
b. Download the dataset JKweather.csv and place it in the main folder
c. In the code file change the name of the station in the following line to :Jammu, Leh ,Sgr to get results for different stations 
dfs = dataset[dataset['Station name'] == 'Sgr']

5. Requirements
The following dependencies are required:
Python 3.8+
TensorFlow 2.x
NumPy
Pandas
Scikit-learn
Matplotlib

6. Methodology
Data Preprocessing: Handling missing values, normalization, and feature engineering.
Model Training: Building and training RNN and LSTM models using time-series data.
Evaluation: Comparing model performance using appropriate error metrics.
7. Citations: NA
8. License & Contribution Guidelines: NA