Swarm-Based LSTM for Temperature Forecasting Using Big Data Analytics

Over the previous few periods, global warming, driven by contamination, the emission of conservatory gases, and the use of non-environmentally friendly dynamism sources, has become a pressing issue. This phenomenon is causing a steady increase in the Earth's overall temperature, which is having profound and detrimental effects on both human populations and ecosystems. Millions of people are at risk of losing their livelihoods due to these changes. One significant consequence of global warming is the escalation of high temperatures during the summer months, resulting in heat-related illnesses and even premature fatalities. To mitigate these risks and better understand climate patterns, researchers have turned to machine learning techniques. Specifically, they have employed algorithms like autoregressive combined poignant averages, collaborative learning, and extensive short-term remembrance networks to forecast temperature trends. This article introduces a novel approach known as CPSO to optimize the hyper parameters within Long Short-Term Remembrance networks. These hyper parameters are fine-tuned by minimizing the mean squared error during the validation of LSTM models. The resulting optimized LSTM hyper parameters are then used to predict temperatures in Chennai, a city heavily impacted by rising temperatures. The CPSO-LSTM model's effectiveness is assessed using a publicly available 25-year dataset of temperatures in Chennai. The evaluation is conducted using MATLABR2020a, with the mean squared error and mean absolute error serving as key performance metrics for the temperature predictions. Significantly, the CPSO-LSTM model outperforms traditional LSTM algorithms in several aspects. Notably, it significantly reduces training time to just 25 minutes while using 200 training epochs and 150 hidden neurons. Moreover, it achieves a lower RMSE of 0.250, as opposed to the outdated LSTM's RMSE of 0.35. This improvement in accuracy is promising, as it can have a positive impact on various applications, ranging from climate modelling to public health planning.