To prevent a collision of a movable part without generating an alarm excessively in manual feed. A machine learning device includes: a state observation unit that acquires, as input data, manual feed state information including a manual feed pulse waveform at a time of a manual feed operation in any manual feed operation performed in a machine tool capable of manual feed; a label acquisition unit that acquires label data indicating a distance by which a movable part of the machine tool moved within a predetermined time immediately after the manual feed pulse waveform of the manual feed state information included in the input data; and a learning unit that executes supervised learning by using the input data acquired by the state observation unit and the label data acquired by the label acquisition unit, and generates a learned model.