Merge branch 'main' of https://github.com/NeuraXenetica/comport_ai

README.md

@@ -5,6 +5,8 @@ Rather than mechanistically deriving the predicted ceiling and floor of a worker
 
 Comport_AI utilizes TensorFlow, Keras, scikit-learn, FastAPI, Uvicorn, Jinja2, NumPy, Pandas, and Matplotlib. It’s developed by Matthew E. Gladden (with support from Cognitive Firewall LLC and NeuraXenetica LLC) and is made available for use under GNU General Public License Version 3. Please see https://www.gnu.org/licenses/gpl-3.0.html.
 
+![Plots exported from Comport_AI](https://user-images.githubusercontent.com/96926263/223198168-18856bb1-f7f6-4019-80a2-111cc3fe6194.jpg)
+
 ___
 ## Steps in the range-modelling process
 The process used by Comport_AI for generating optimized models for the likely range of workers’ future job performance involves the following steps.
@@ -17,6 +19,9 @@ It’s possible for a user to configure a number of the simulation’s parameter
 The raw data that will be used for training models and predicting the range of workers’ future performance takes the form of a synthetic dataset of factory workers’ daily behaviors generated by the open-source Synaptans WorkforceSim platform. Comport_AI’s web interface allows a user to either upload a WorkforceSim file (of the type “wfs_behaviors_and_records_[*].csv”) or choose from one of two WorkforceSim datasets built into Comport_AI: a “Small” dataset with roughly 20,000 observations or “Medium” dataset with ≈119,000 observations. Building models based on the Medium dataset may take a very long time or exceed the memory available on many computers; it’s thus recommended that one begin with the Small dataset to test the app’s functioning before attempting to use a larger dataset.
 #### Selecting the modelling settings
 A user can choose to predict either the mean value of the daily efficacy scores that will be registered for a worker in a given future time period or the number of “teamwork” behaviors that will be recorded for the worker during that period. The available time periods are the next 7 days or the next 30 days. (The modelling of workers’ registered daily efficacy scores for the next 30 days is recommended as an initial experiment, as it involves the richest dynamics and makes it easiest to visually compare different approaches to the modelling of workers’ performance ranges.) A user can also specify a random seed, in order to facilitate reproducibility of results.
+
+![The Comport_AI web app interface](https://user-images.githubusercontent.com/96926263/223198707-776b2164-f7ff-4853-b7e9-3c3e1439e9dc.jpg)
+
 ### Training the performance-range models
 When a user clicks on the green button to “train the performance-range models and compare their results,” the package begins working to carry out the following steps.
 #### Producing the Base Target Model
@@ -43,6 +48,8 @@ While the metrics for Ceiling and Floor Models previously described were useful
 - **Mean Summed Absolute Distances to Range Edges (MSADRE).** This is sum of the absolute distance of an actual target value from its predicted ceiling value and its absolute distance from its predicted floor value, averaged across all cases.
 - **Overall Range Performance (ORP).** For a given Joint Range Model, this complex metric equals `(1 – PATOR)² ÷ √(MSADRE)`. It more significantly penalizes models that have a greater share of actual target values falling outside their predicted range, while only relatively weakly penalizing models that yield larger predicted ranges. Unlike in the cases of OCE and OFE, our goal in the modelling of performance ranges is to maximize the value of ORP.
 
+![Plots exported from Comport_AI](https://user-images.githubusercontent.com/96926263/223199212-15568690-ee0d-4e04-9499-367e787ac9d1.jpg)
+
 ___
 ## Evaluating the results
 Once the process of producing all of the Joint Range Models has been completed, Comport_AI will display plots visualizing the performance of the Base Target Model and Ceiling, Floor, and Joint Range Models. Each plot includes data on the given model’s metrics; it’s also possible to use the green button that will appear at the bottom of the webpage to download a CSV file with all of the model metrics.