Optimizing Hidden Markov Models for the Internet of Things

Sammanfattning: The Internet of Things is a new and emerging branch of control systems where gatheredsensor data is analyzed using computer algorithms. This report applies Markov models, astatistical model based on states with transitions between them, to temperature datagathered from an office. In particular, hidden Markov models are used, which infer the statesand their probable transitions from the temperature observations. The purpose of this studyis to find the optimal number of states in a hidden Markov model for Internet of Thingstemperature data. Evaluation is made by comparing the difference between data predictedby the Markov model and actual temperature measurements. 10, 20, 24, 30, 40, 50 and 168states are evaluated. All number of states perform poorly and differ greatly from the actualmeasurements. It is concluded that a more thorough study is needed to accurately answerthe problem statement.