Bacterial-based self-mending concrete (BSHC) has been perceived for its exceptional break recuperating limit, however it is frequently exorbitant and tedious, restricting its application to lab settings. AI (ML) models offer an answer by foreseeing the mending execution (HP) of BSHC, consequently saving time and expenses related with lab tests, microscopic organisms’ determination, and recuperating system reception. Ureolytic bacterial healing concrete (UBHC), aerobic bacterial healing concrete (ABHC), and nitrifying bacterial healing concrete (NBHC) are the three types of BSHC that are the subject of this investigation. Support vector regression, decision tree regression, deep neural network, gradient boosting regression, and random forest are the five ML algorithms used. These models employ 22 influencing factors as variables. A dataset of 797 BSHC tests from writing (2000-2021) approves the ML models. The grid search algorithm (GSA) is used for parameter tuning. The coefficient of determination (R2) and the root mean square error (RMSE) are used to evaluate the performance of the model. The GBR model exhibits unrivaled forecast capacity (R² = 0.956, RMSE = 6.756%) contrasted with different models. Awareness examination inside the GBR model further looks at the effect of factors on HP.