Rain or Shine? An Exploration of Modern Rain Forecasting project presentation

This presentation delves into the science behind rain forecasting, exploring various techniques used to predict precipitation patterns. Learn how meteorologists use data, models, and technology to forecast rain and make informed decisions. for more details visit: https://bostoninstituteofanalytics.org/data-science-and-artificial-intelligence/

1. Rain Forecasting Ruchita Patil

2. Agenda Training a machine learning model trained on historical rainfall data to give accurate predictions for upcoming months. Visualizaising and explaining past rainfall data and displaying forecasted rainfall for upcoming months. Reviewing model training process, performance evaluation metrics, and limitations.

3. Analysis Used line plot to see the trend line of rain over the years

4. Rainy season analysis Trends Over Time: Early 1900s (1900-1920): The data shows moderate variability with some high peaks around 1920. 1920-1960: This period exhibits high variability with significant peaks and troughs, particularly around the 1940s and 1950s. 1960-1980: The plot indicates a smoother trend with moderate peaks, suggesting relatively stable rainy seasons during these years. 1980-2000: There is an increase in variability again, with more pronounced peaks and troughs. 2000-2020: The most recent data shows increasing peaks, indicating potentially higher rainfall amounts or more frequent rainy days.

5. Non rainy season Trends Over Time: Early 1900s (1900-1920): There are very high spikes, indicating some years had significantly long or frequent non-rainy seasons during this period. 1920-1960: This period shows more variability but with lower peaks compared to the early 1900s. The frequency and intensity of non-rainy seasons are reduced. 1960-1980: There’s a noticeable dip with relatively fewer and lower spikes, suggesting fewer significant non-rainy seasons. 1980-2000: The spikes start to become more prominent again, indicating a potential increase in non-rainy season events. 2000-2020: The spikes return to higher values similar to the early 1900s, showing an increase in non-rainy season lengths or frequency.

6. Model Training Performed AD fuller test to check if the data is stationary.

7. Model Training Rainfall data often exhibits clear seasonal patterns due to natural climatic cycles. These cyclical trends make rainfall data well-suited for models that can handle seasonality, so i used Sarimax model.

8. Prediction using Sarimax

9. Alternate Model Used - LSTM I also used LSTM model for time series but it didnt show me promising result so i sticked to the sarimax.

10. Thank You!