Abstract The quality of water resources and aquatic life as well as irrigation projects are affected by suspended sediment loads (SSL) in rivers. Therefore, it is important to study SSL prediction with the aim of observing it to mitigate losses and risks. This study proposed and tested two algorithms related with deep and machine learning (ML) are Long Term Memory (LSTM) and Random Forest (RF) to predict monthly SSL in Greater Zab River tributary catchment. This dataset contained monthly discharge data 𝐃 for the period of 1981– 2008. Seven scenarios including different inputs are used to find the best model. The performance of these models is compared using several metrics. The results of the experiments showed the superiority of the LSTM algorithm over RF algorithm. Through the analytical study, the best performance of the LSTM model is determined to predict SSL in scenario 6 (input is current month-discharge 𝐃𝐭 , previous month-discharge 𝐃𝐭−𝟏, previous two months-discharge 𝐃𝐭−𝟐, previous month-sediment 𝐒𝐭−𝟏), where it obtains the highest coefficient of determination (𝐑 𝟐= 0.861) and the lowest mean absolute error (𝐌𝐀𝐄 = 0.020) and root mean square error (𝐑𝐌𝐒𝐄 = 0.044) compared to the other models.