Microchannel heat sinks are receiving attention for thermal management of datacenter and electronics cooling applications. Because the parallel microchannel cooling heat sinks (PMCHS) have high heat transfer coefficients and are easy to fabricate compared to other complex heat sinks. Similarly, the oblique fin-pin heat sinks also have simple geometry and exhibit a slightly enhanced heat transfer coefficient compared to the straight channels. Many studies have been performed to understand the thermal performance of straight and oblique heat sinks. This study aims to establish a comprehensive machinelearning model capable of predicting the Nusselt number for straight channel and oblique fin-pin heat sinks. The models utilized a database of 893 data points from 5 separate studies. Totally 10 input features are used in the training and testing dataset. In this study, four distinct machine learning algorithms were considered: random forest, LightGBM, XGBoost, and KNN models. The performance of all four models was compared with R-squared, mean squared error (MSE) and mean absolute error (MAE). The machine learning models obtained the MAEs of 0.75 to 1.72 and represented fivefold model predictions accuracy. However conventional regression correlation models make it difficult to predict the high accuracy of the Nusselt number.