EGH_Extension

In the past, I have used a model, called "Mixture model of EGHs" to predict the next activity ahead. The idea is to analyse the past occurrences of the target event that we would like to predict. By looking the preceding window of the past occurrences, we can create a single sequence of these preceding sequences. The statistical mixture model could be generate to summarise this training dataset, so we can use this model to predict whether or not the target event would occur in the next time slot, given the recent window sequence of events.
The original algorithm of mixture model of EGHs considers all the occurrences of the events in the past. By applying to the home energy management scenario, we have observed that the day of the week is one of the important factor for the homeowners to decide which appliances/events/activities they could use. For example, the activities are obviously different between weekdays and weekends. This is because the homeowner' behaviours can be impacted by others external factors such as local weather, working hours, social activities.
Thus, taking the weekday as an important factor to predict the future events/activities, we can extend EGH algorithm. Instead of computing all past occurrences of the target event (i.e., the event that we would like to predict), we only consider the past occurrences that occurred to the same weekday that we would want to predict.
For example, if we want to predict occurrences of "washing machine" on next Tuesday, we only analyse the past preceding sequences that had the "washing machine" running on Tuesday.
By doing so, we might catch the patterns between appliances/activities and improve the prediction.
As the human's routines are best described daily, and weekly. The size of preceding window should be 7 days (1 week).

I have finished the code. Debugging at the moment...