International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 12 Issue: 04 | Apr 2025
p-ISSN: 2395-0072
www.irjet.net
Improving Weather Forecast Accuracy in India using Synthetic Data Generation Neil Chaudhary Student, Step By Step School, Delhi, India ---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - This study examines the influence of synthetic
Even with improvements in weather forecasting models, there remains a wide gap in India's capacity to predict weather with the accuracy needed to counter the risks posed by unpredictable climate patterns. Although machine learning algorithms have been promising in enhancing forecasting accuracy, the absence of highdensity data coverage is still a challenge. Further, not much research has been done on how the inclusion of synthetic data can improve weather prediction accuracy in countries such as India.
data generation on the accuracy of weather prediction in India. From the Indian Weather Repository dataset, we used a Tabular Generative Adversarial Network (TabGAN) to generate synthetic weather data points and measured the influence of these points on machine learning regression model performance. Temperature prediction in Celsius was the focus of the dependent variable. Findings show that incorporating artificial data had a substantial increase in predictive accuracy over baseline models learning from the initial data alone. The method provides a likely means of overcoming the shortfall created by India's thin network of weather monitoring stations, which could be advantageous for key sectors like agriculture, disaster management, and planning resources.
Better prediction of weather patterns in India has extensive implications. Tens of millions of farmers rely on precise and timely forecasts of weather patterns to schedule their agricultural operations and hence it becomes critical to dampen the risk of unpredictable weather. Better performance of weather forecast models could dramatically contribute to lower crop losses and higher food security in India. The economic dividend can go much beyond agriculture to influence disaster mitigation, water resources planning, and even power generation.
Key Words: Synthetic Data Generation, Weather Prediction, Tabular Generative Adversarial Network, Machine Learning, Regression, Temperature Prediction, Generative Adversarial Network, Artificial Data
1.INTRODUCTION
In this research, we tested if it was possible to enhance the accuracy of prediction using a weather dataset by including more data points. The Indian Weather Repository was used as the dataset for this research, and the dependent variable was temperature in Celsius. Synthetic data was created through the use of a Generative Adversarial Network (GAN), which was then added to the original dataset for training and testing of machine learning regression models.
Weather forecasting is a complex process that requires monitoring several environmental variables to make predictions about future weather conditions with accuracy. But the prediction's accuracy differs greatly from one region to another, with nations like India having more difficulties compared to Western countries. The reason for this difference mainly lies in the sparsity of observation stations of weather in India, which does not allow the acquisition of the required amount of data to make accurate forecasts.
2. RELATED WORK The application of synthetic data towards enhancing predictive models has been widely explored across a number of disciplines, including meteorology. Such research has provided evidence of data augmentation methods helping to enhance performance in models where
In contrast to European and North American nations with dense weather station networks, India has a small number of data points, and hence less accurate predictions, particularly for localized phenomena like monsoons or heatwaves. Since India is so geographically diverse and weather forecasting is so important for key sectors like agriculture, it is essential to enhance prediction accuracy. Farmers, especially, depend upon reliable weather forecasting to make important decisions concerning planting, irrigation, and harvesting of crops.
© 2025, IRJET
|
Impact Factor value: 8.315
data collection platforms are limited, especially in such areas. According to McMurdie et al. (2019)[1], weather prediction accuracy in areas with dense data is assured by the simple fact that locations with higher density of weather stations tend to show greater forecast accuracy.
|
ISO 9001:2008 Certified Journal
|
Page 145