This paper presents the streamflow modelling of River Niger at Lokoja and Onitsha streamflow gauging station in Nigeria based on least square regression technique. The flow calibration was performed for eleven years from 1998 to 2008 for River Niger at Lokoja streamflow gauging station and from 1980 to 1990 for River Niger at Onitsha streamflow gauging station. The models were also verified and they gave good results. The performance of the modelling effort was evaluated by calculating statistical measures such as Coefficient of Determination (R2) ranging from 0.9983 and 0.9995, Coefficient of Correlation (R) ranging from 0.9991 and 0.9997, Nash-Sutcliffe model Efficiency (NSE) ranging from 0.9962 and 0.99996, Root Mean Square Error (RMSE) or Standard Error of Estimate (Syx) ranging from 12.68 and 35.66, Mean of Residues (MR) or Mean Absolute Error (MAE) ranging from 7.042 and 33.03 and Mean Absolute Relative Error (MARE) ranging from 0.9981 and 0.9996 and then the measured (observed) state variables are compared with the predicted (simulated) state variables. The developed models can be used for hydropower electricity generation, water resources development and management within Lokoja and Onitsha towns and to study any post development scenario within the River Niger catchments and for further hydrological studies in the river basin.