Soil moisture content improvement is a key process in agricultural production and food security in arid and semi-arid regions. The interaction effect of crop water requirement (CWR) and soil texture on water productivity (WP) was evaluated in Harbin, Heilongjiang province, China. A field experiment with two scenarios of compost application (0, 25 ton/ha) and three levels of irrigation policies (deficit irrigation (DI = 0.75CWR), regular irrigation (RI = CWR), and full irrigation (FI = 1.25CWR)) was planned in three replications during the 2021–2022 year. Water productivity was simulated as a criterion to improve the irrigation time for increasing the final biomass. Four strategic crops including potato, corn, wheat, and barley were incorporated into the daily simulation system. Furthermore, a machine-learning random forest algorithm was used to find the best irrigation times. The results showed that the use of adjusted irrigation time and improved soil texture can increase water productivity by reducing evaporation and deep percolation and increasing actual biomass. Estimation of irrigation time in a learning-based method will be optimal when plant growth and soil moisture are monitored on a daily basis.