Right now, the dashboard does not display plots from the generated .csv files. To fix this issue, I rewrote old plotting functions from javascript d3 to python plotly, which easier to understand/maintain.
Right now, the function dashboard_plot.py can plot:
Stacked bar charts
Bar charts
Area charts
In the future when we have more data, I will add more types of plots to the function.
Display model accuracy
To display the model accuracy, I first get a list of built models in the ModelContainers. Then, I average the scores (accuracies, classification balanced rates, and Matthews coefficients) across the models and display them on the dashboard (top left corner, below the number of experiments).
Note: There is the class StatsModel; however, I don't think this model contains the information that we needed. Once Rui is done with his rec-system, we might have to discuss a systematic way to store the model data, and how to get them.
Screenshot
Finally, here is the screenshot for the dashboard right now in my branch. Please comment below in this PR thread over the break; I will be sure to get back to you!
Fix dashboard plots
Right now, the dashboard does not display plots from the generated
.csvfiles. To fix this issue, I rewrote old plotting functions fromjavascript d3topython plotly, which easier to understand/maintain.Right now, the function
dashboard_plot.pycan plot:In the future when we have more data, I will add more types of plots to the function.
Display model accuracy
To display the model accuracy, I first get a list of built models in the
ModelContainers. Then, I average the scores (accuracies, classification balanced rates, and Matthews coefficients) across the models and display them on the dashboard (top left corner, below the number of experiments).Note: There is the
class StatsModel; however, I don't think this model contains the information that we needed. Once Rui is done with his rec-system, we might have to discuss a systematic way to store the model data, and how to get them.Screenshot
Finally, here is the screenshot for the dashboard right now in my branch. Please comment below in this PR thread over the break; I will be sure to get back to you!