Have a personal or library account? Click to login
Unicorn, Hare, or Tortoise? Using Machine Learning to Predict Working Memory Training Performance Cover

Unicorn, Hare, or Tortoise? Using Machine Learning to Predict Working Memory Training Performance

Open Access
|Sep 2023

Figures & Tables

joc-6-1-319-g1.png
Figure 1

General Flowchart of Analytic Plan.

Note: Three colors represent different groups identified by clustering.

joc-6-1-319-g2.png
Figure 2

Learning Patterns Identified by Clustering.

Note: A) Scatter plot of participant’s optimal knot location (x axis) and initial training gain (slope1; y axis) as a function of the three learning clusters. B) Average learning trajectories of the three clusters across the 10 training sessions. Error bars represent ±1 standard error of the mean.

Table 1

Confusion Matrix of Binary Tree Model Performance on Cross Validation Set.

MODEL1 (N = 377)PREDICTED CLASS
UNICORNNON-UNICORN
True classUnicorn28 (.88)4 (.12)
Non-Unicorn62 (.24)283 (.82)
MODEL2 (N = 345)TORTOISEHARE
True classTortoise114 (.64)63 (.37)
Hare62 (.36)106 (.63)
Table 2

Confusion Matrix of Binary Tree Model Performance on Test Set.

MODEL1 (N = 95)PREDICTED CLASS
UNICORNNON-UNICORN
True classUnicorn5 (.63)3 (.37)
Non-Unicorn14 (.16)73 (.84)
MODEL 2 (N = 76)TORTOISEHARE
True classTortoise21 (.57)16 (.43)
Hare14 (.39)22 (.61)
Unicorn2 (.67)1 (.33)
joc-6-1-319-g3.png
Figure 3

Predictor Importance by Calculating the SHAP Value.

Note: The x-axis represents the SHAP value; The y-axis represents the predictors. Each dot represents each participant who is correctly predicted by our model (including both hitting a particular class and correctly rejecting the class), the color of the dot represents a certain predictor’s value (red: higher value, blue: lower value). A. Each predictor’s SHAP value in predicting the unicorn group (model 1) ranking by predictor importance. B. Each predictor’s SHAP value in predicting the tortoise group (model 2) ranking by predictor importance.

joc-6-1-319-g4.png
Figure 4

Scatter Plots of Predictor Values (x axis) and SHAP Values (y axis).

Note: All predictor values were transformed to z scores. The top two important predictors in predicting unicorns (green) versus non-unicorns (gray) (Figure 4A) and tortoises (orange) versus hares (purple) (Figure 4B) are shown.

DOI: https://doi.org/10.5334/joc.319 | Journal eISSN: 2514-4820
Language: English
Submitted on: Feb 16, 2023
|
Accepted on: Aug 15, 2023
|
Published on: Sep 4, 2023
Published by: Ubiquity Press
In partnership with: Paradigm Publishing Services
Publication frequency: 1 issue per year

© 2023 Yi Feng, Anja Pahor, Aaron R. Seitz, Dennis L. Barbour, Susanne M. Jaeggi, published by Ubiquity Press
This work is licensed under the Creative Commons Attribution 4.0 License.