A sump pump system may implement adaptive learning and machine learning techniques to facilitate improved control of sump pumps. A sump pump system may implement the described techniques to generate, train, and/or implement a machine learning model that is capable of predicting or estimating one or more conditions of the sump pump system (e.g., water level in the basin, motor malfunction, stuck impeller, geyser effect, blocked outlet pipe, faulty level sensor/switch, faulty bearing, failure to engage pump at high-water mark, etc.) based on one or more detected input variables (e.g., acceleration or vibration patterns detected in water, on a pump, or on a pipe; capacitance values of water; audio signatures; electrical signatures, such as power or current draw; pump motor rotation speed; water pressure signatures or values, such as those detected at the bottom of a sump basin; etc.).