<p>The tropical cyclones (TC) are huge, catastrophic, and destructive natural disasters cause coastal flooding, damage infrastructure and forcing people to evacuate. Previously, little or no TC warning existed due to the unavailability of latest instruments and modern weather-related studies. However, due to recent technological advancements, it is possible to provide early warning and precautionary measures in advance. This study analyses the causative relationship between solar activity and TC genesis in the Pakistan and Oman region, the North Indian Ocean, using machine learning modelling approaches, including adaptive neural-fuzzy inference system (ANFIS), Bayesian Regularized Neural Network (BRNN) and Random Forest (RF). For the selection of appropriate solar fluences (proton, electron, electron density, speed, temperature and the ratio of alpha particles), the principal component analysis is applied. It is found that ANFIS model outperformed overall and specifically with proton fluences (P1MeV, P10MeV, P100MeV) and electron fluence (E0.6MeV) as input parameters based on root mean square error values. Moreover, precision, recall and F1 score suggest that almost all the models may capture up to 33% of the true calculated TC. Furthermore, the sensitivity analysis of aforementioned parameters, performed using the nonlinear exponential model, suggests that P100Mev (lag 1) may have a strong impact on TC genesis. One of the limitations of this study is analysing input data up to a maximum of 5 lags due to computational cost of the model. The results presented in this study may be a good contribution to the relationship of TC with solar activities and its predictability to minimize future adverse impacts in the Asian region.</p>