A Future Search Optimization based Hybrid Meta-Heuristic Algorithm for Pandemic Spread and Mortality Prediction

In the complex world of public health, predicting the future is often a matter of life and death. When a virus spreads, governments need to know not just how many people are sick today, but how many might fall ill tomorrow. This is the realm of time-series modeling, a method of looking at past patterns to forecast what comes next. For decades, scientists have used statistical tools to draw lines through data points, hoping to see the shape of things to come. However, the spread of disease is rarely a straight line; it is a chaotic, shifting phenomenon influenced by human behavior, policy changes, and biological factors. To navigate this complexity, researchers have turned to artificial intelligence, specifically a type of computer program called a neural network. Think of a neural network as a digital brain that learns from examples, adjusting its internal connections until it can recognize patterns that traditional math might miss. Yet, even these digital brains can get stuck or make mistakes if they are not guided correctly. This is where a newer class of problem-solving tools, known as meta-heuristic algorithms, enters the picture. These are computational strategies inspired by nature—such as how birds flock or how whales hunt—that help computers search through vast possibilities to find the best possible solution, avoiding dead ends and local traps.

Against this backdrop of urgent need and technological innovation, a researcher at Khatam University in Tehran set out to refine how we predict the deadly toll of the COVID-19 pandemic. Focusing on the Netherlands, a country that experienced significant waves of infection, the study analyzed daily death records from the beginning of the outbreak in February 2020 through April 2021. The goal was not merely to count the dead, but to build a more accurate forecasting machine. The researcher began by feeding the computer a wide array of data points derived from the daily death numbers. Instead of just using the raw count of deaths, the team transformed this data into sixteen different mathematical indicators, similar to how a financial analyst might look at moving averages or momentum to predict stock prices. These indicators helped the computer see the trends, the speed of change, and the volatility of the pandemic's curve. To ensure the computer was not overwhelmed by unnecessary information, the researcher employed a genetic algorithm, a process that mimics biological evolution, to sift through the data and select only the most relevant indicators for the final model.

With the data prepared, the study tested several different ways to train the prediction model. The baseline was a standard artificial neural network, a powerful tool on its own. However, the researcher then introduced four different meta-heuristic algorithms to see if they could guide the neural network to a better answer. These included the Moth-Flame Optimization, which simulates how moths navigate by light; the Whale Optimization Algorithm, which models the bubble-net hunting strategy of humpback whales; the Giza Pyramids Construction algorithm, which draws on the physics of moving heavy stones up a ramp; and a newer method called Future Search Optimization. This final algorithm is unique in its approach, modeling how people in society try to improve their lives by emulating the most successful individuals, constantly adjusting their path toward a global ideal. The computer was asked to run these different strategies thousands of times, each time trying to minimize the error between its prediction and the actual historical data.

The results of this extensive digital experiment were clear and decisive. While the standard neural network provided a decent forecast, it was outperformed by the algorithms that used nature-inspired guidance. Among the various methods tested, the Future Search Optimization algorithm emerged as the most precise tool. It achieved a level of accuracy that was orders of magnitude better than the other methods, including the whale and moth-inspired strategies. The study found that this algorithm could navigate the chaotic data of the pandemic with a degree of precision that the other models could not match, effectively finding the optimal path through the noise of daily fluctuations. The research demonstrated that by combining the pattern-recognition power of neural networks with the strategic searching ability of these advanced algorithms, it is possible to create a model that is both faster and more accurate. The study concludes that for future pandemics, relying on these hybrid methods could provide governments with the reliable foresight needed to make timely decisions, potentially saving lives by allowing for earlier and more effective interventions. The work suggests that the key to understanding the future of disease spread lies not just in looking at the past, but in using the right computational tools to interpret it.