Activity prediction using context

• Main Authors: Co, Janelyn, Samson, Cheney, Uy, Justin
• Format: text
• Language: English
• Published: Animo Repository 2011
• Online Access: https://animorepository.dlsu.edu.ph/etd_bachelors/11823
• Institution: De La Salle University

As new technologies emerge, it is important to take note of the possible improvements we can make in terms of how humans and computers interact. Construction of an empathic environment that supports and responds to the users state is becoming more of a necessity for intelligent systems. Recognizing and predicting human activities will provide these intelligent systems the means to properly assess the needs of its users. This research make use of previous actions and context containing item used, location, and affect to predict the possible future activities that the user might do. A corpus that includes affect and actions of the user will be used to model the activities of the inhabitant. The model is able to anticipate the succeeding activity of the user. The proponents have tested 4 types of model on 2 different DBN set-ups. The models that were used in testing were Item used only, item used with location, item used with emotion, and item used with location and emotion. With the different combination of attributes tested the proponents have obtain these results. For set-up 1 item used had 32%-56%, item used with location had 33%-56%, item used with emotion had 32%-37%, and item used with location and emotion had 33%-56%. For set-up 2 item used had 33%-61%, item used with location had 34%-56%, item used with emotion had 34%-61%, and item used with location and emotion had 37%-56%. The results show that set-up 2 (refer to Figure 4.5) has a much higher output that set-up 1 (refer to Figure 4.4), this is because of how the set-ups were designed.