Microlocations using tagged data

A method of predicting actions for a mobile device, the method comprising performing by the mobile device:

• collecting a tagged sample set by, for each respective tagged sample of a plurality of tagged samples in the tagged sample set, the tagged sample set including a respective first data point that includes one or more sensor values from one or more signals emitted by one or more signal sources located within a physical area, a respective action of a plurality of actions, and a respective label representing the respective action at a respective target location of a plurality of target locations;
• collecting an untagged sample set, each respective untagged sample of a plurality of untagged samples in the untagged sample set including a respective second data point, the respective second data point including one or more sensor values from one or more signals emitted by the one or more signal sources and measured by the mobile device using one or more sensors;
• training a semi-supervised machine learning model using the plurality of tagged samples in the tagged sample set and the plurality of untagged samples in the untagged sample set, the semi-supervised machine learning model being trained to classify an input data point as corresponding to a particular target location of the plurality of target locations;
• obtaining a current data point by measuring one or more sensor values;
• inputting the current data point to the semi-supervised machine learning model to obtain a predicted target location, the predicted target location corresponding to a predicted action of the plurality of actions;
• and
• performing or providing the predicted action.