Understanding flow dynamics is essential for effective water resource management, particularly in regions prone to hydrological variability and extreme weather events. The overarching question of our study is to unravel the complexities of river flow prediction using machine learning algorithms. We focused on daily flow prediction in the Upper Humber River Catchment (UHRC) to answer this effectively. The study identified important variables in predicting daily river flow and potential algorithms to enhance flow prediction models, thereby improving integrated water management practices. Our methodology involves extracting daily flow records from Environment Canada from 1965 to 2022 across five stations within the UHRC: Cold Creek, Elder-mills, East-Humber, Pinegrove, and West-Humber. Data preprocessing, feature engineering, and analysis were carried out within the R ecosystem using CARET. For data preprocessing, missing values were removed, and feature extraction such as discharge lags (1-day, 2-days, and 3-days) and various ratios were determined and integrated into the dataset. Regression algorithms such as random forest, multivariate adaptive regression splines, support vector machine, and K-nearest neighbour were applied to the study. Notably, 1-day lag, the sum of runoff, 2-day lag, the sum of surface runoff, and the ratio of air temperature to lake temperature emerge as important variables across these algorithms, with 1-day lag being the most influential. The importance of lag in flow prediction provides valuable insight into the temporal dynamics of river flow within the UHRC. Specifically, 1-day and 2-day lag suggests a short-term temporal dependency in the river flow dynamics within the UHRC.