Analyze temperature and atmospheric pressure in a residential space
Do the following activities using Python and answer the questions of the last bullet. The exam will be done and delivered in teams.
-
Analize the data file ds-home-5min-avg.csv the document. In this file are recorded measurements of humidity (hum), temperature (tem) and atmospheric pressure (pre) in a given datetime (date), along with a discrete occupancy level (occ), for a residential space (living room + dinner room).
- Generate plots for describing the statistical distribution of every feature.
- Calculate and plot the correlation matrix between numerical vaues (hum, tem, pre).
-
Train three data models for predicting occupancy levels (occ) given humidity, temperature and pressure readings. Document the parameters used that produce the best result for each model. Models can be:
- Decision Trees
- SVM
- Neural Nets
- KNN
- Other
-
Evaluate the trained models using the metrics below. Document the configuration used for making the evaluation (eg. training-test partition).
- Accuracy
- F-Score
- Precision y Recall
- Confussion Matrix
-
Answer the following questions and justify your answers:
- Which was the best model and which metric did you used for choosing it?
- Which kind of errors were the most common for the chosen classifier?
- Which other characteristics could be generated from current data in order to improve your model?
-
Document your results and submit them through this assignment before the due date indicated below. Use tables to summarize and contrast your results. Attach the following files:
- A PDF document with your results
- The python script or Jupyter notebook with your experiments.
5 extra-points will be granted to the team that improves the model accuracy by incorporating additional features to the model.