ESTIMATING DISTRICT-LEVEL GDP FROM NIGHT-TIME LIGHT DATA USING DEEP LEARNING TECHNIQUES

Abstract

Gross Domestic Product (GDP) is a fundamental indicator reflecting the economic performance and overall health of a nation's economy. In recent years, India has experienced significant growth in urban areas, characterized by diverse economic activities. This development underscores the need to accurately calculate the Gross District Domestic Product (GDDP). In developing countries like India, accurately deriving GDP poses challenges due to unrecorded factors that significantly contribute to the country's economic activities. Traditional methods of GDP calculation are not only expensive in terms of time and resources but also prone to biases. This study proposes an innovative approach to efficiently measure GDDP using night-time light data from the VIIRS satellite and deep learning techniques. By leveraging night-time light data to analyze economic activity, the estimation process becomes independent of the informal sector's unique characteristics. The research focuses on estimating GDDP for districts within India's states by utilizing the spatial information from night-time light data. Deep learning techniques such as Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), Recurrent Neural Network (RNN), and hybrid models combining LSTM and GRU have demonstrated promising results in handling sequential and time series data. The results indicate that deep learning methods can effectively serve as proxies for estimating district-level GDDP using night-time light data. The study concludes by discussing the implications of these results for GDDP estimation, emphasizing the potential of this approach to provide timely and more reliable estimates. This research paves the way for enhancing GDDP estimation using night-time light data, offering a more efficient and dependable methodology.