An advanced framework for predicting pollen outbreaks using multi-input multi-output temporal convolutional networks and firefly algorithm.

Pollen outbreaks pose serious public health concerns, especially for those with respiratory problems. Reliable detection systems, which provide timely interventions, are essential to mitigate these risks. This work presents novel pollen outbreak prediction by integrating Multi-Input Multi-Output Temporal Convolutional Networks (MIMO-TCN) with Firefly algorithm for optimal hyperparameter tuning. It presents a technique that uses the Firefly Algorithm (FA) for hyperparameter tweaking in conjunction with the Multi-Input Multi-Output Temporal Convolutional Networks (MIMO-TCN) architecture. The FA, inspired by firefly communication and mating habits, excels at optimization challenges. We ran extensive trials on a dataset of 20,000 to 100,000 instances, testing the model using measures including accuracy, precision, sensitivity, and F1 score. The model beat various state-of-the-art algorithms, with an average precision of 0.965, sensitivity of 0.982, F-measure of 0.973 and accuracy of 92.04%. Confusion matrix analysis demonstrated strong performance across pollen types (Tree, Grass, and Weed), emphasizing the model's classification accuracy and real-world applications.