This week, I am working to get the dynamic time warping functionality into my program. The process of doing so includes re-processing the features to include the time series, putting each series back together when we construct sequences, and then performing the DTW to generate a number that will be used to compute the kNN of each sequence which can then be used for predictions with the models. The processing time of these activities has gone up significantly since we have been using five different metrics with each of the F phase datasets. I am returning to school next week, and once I've completed the DTW processing all that will remain before we put together our second paper (The date for the reach journal we would like to submit it to is October 1), I am hoping I will have time to look again into the Agglomerative Hierarchical Clustering concept, which I did not successfully complete when we explored it earlier in the summer and then changed focus to the paper. We heard back
To summarize the results of my change analysis exploration, I decided to graph the changes in three positive and three negative attributes for each patient and for each comparison (baseline, the previous day, average of all previous days). The resulting charts were quite busy, but there were a few observations that may be helpful going forward:
- most patients had positive outcomes across my chosen attributes when comparing against baseline
- average of all previous days may be a helpful comparison to make since it can give a measure of past data without being affected by day-to-day variability, which can be quite dramatic
Based on feedback from Dr. Sprint and Alexa, I am going to re-format the line graphs into bar charts and remove the previous day comparison altogether.
I am not discouraged completely by the lack of a solid conclusion based on this exploration, because the actuality of the situation is that I chose the features without a real, solid reason as to why they would be the correct ones to look at. My objective was to visually represent a change in attributes over time, and I think I have effectively done so. Once we have met with Dr. Weeks and Dr. Skornyakov, I am hopeful that they will be able to help us decide which attributes will actually be valuable to look at and I'll have the know-how to model them appropriately.
Now that I have completed this exercise, I am going to begin looking at each lights on's attributes in comparison to the night before and the night following. Since our ultimate goal is to look at a night's relationship with the following day, I will begin by graphing sleep minutes per night against wake minutes during the following day. I also plan to review studies that classify daytime activity to explore how we might go about classifying activity during the day other than looking at raw activity numbers.
Comments
Post a Comment