Machine learning device, screen prediction device, and controller

A machine learning device comprising:

• a state observation unit that acquires a state vector sequence {xn:
• n is equal to or greater than 1} that is time-series data as a group of state vectors xn corresponding to a transition of operation screens responsive to a series of operations by a user or change in a machine state of an industrial machine, the state vectors xn each containing screen type data yn and machine state data zn, the screen type data yn indicating a screen type of each operation screen used by the user when the user performs the series of operations on the basis of an operation screen displayed on a display unit of a controller for controlling the industrial machine, the machine state data zn indicating the machine state of the industrial machine controlled by the controller when the operation screen is displayed;
• a training data acquisition unit that acquires training data containing input data and label data, the input data being a subsequence of preset N (N is equal to or greater than 2) consecutive state vectors {xn, xn+1, . . . xn+N−1} starting from an arbitrary state vector xn (n is equal to or greater than 1) belonging to the state vector sequence acquired by the state observation unit,
• the label data being screen type data about a screen to be displayed next or on an occasion after the next to the operation screen containing screen type data yn+N−1 indicating the type of a screen to be displayed next to an operation screen corresponding to a state vector xn+N−1 at the end of the state vector subsequence as the input data;
• and
• a learning unit that performs supervised machine learning using the training data containing the input data and the label data, when the user performs a series of operations on the basis of an operation screen displayed on the display unit, the learning unit generating a learned model for predicting screen type data to be used next or on an occasion after the next by the user on the basis of a state vector subsequence containing N consecutive state vectors determined retrospectively starting from a current state vector.