ctgame-analytics — demo analysis report

Built automatically by the CI/CD pipeline with ctgame 0.1.0 on synthetic data (no real student data). Source: github.com/beliyaka98/ctgame-analytics

1. Data

120 students, 40 tasks, 3648 attempts; 59% of student-task pairs were tried.

2. Latent skill factors (masked NMF)

3 factors, RMSE on observed cells = 0.3213, 151 iterations, converged: True. On 20% held-out attempts the prediction error is 0.3727 against 0.4376 for the task-mean baseline (15% lower).

Heatmap of NMF task factors

3. Learner profiles (PCA + K-means)

Best k = 3 by silhouette; cluster sizes [41, 40, 39].

Students on two principal components coloured by cluster
success_analysis success_evaluation success_explanation success_inference
cluster
0 0.32 0.75 0.62 0.40
1 0.27 0.53 0.22 0.93
2 0.90 0.13 0.17 0.82

4. Pilot evaluation (ANCOVA)

F(1, 57) = 11.16, p = .001, partial eta^2 = 0.16; adjusted difference = 4.04 [95% CI 1.62, 6.45]. Assumptions met: True.

Adjusted post-test means by group

5. Next-task recommendations (first 5 students)

student_id task_id reason predicted_success observed_success distance
S001 T008 new 0.83 NaN 0.13
S001 T015 new 0.86 NaN 0.16
S001 T027 new 0.90 NaN 0.20
S002 T034 new 0.88 NaN 0.18
S002 T018 new 0.89 NaN 0.19
S002 T030 new 0.90 NaN 0.20
S003 T014 new 0.59 NaN 0.11
S003 T010 new 0.96 NaN 0.26
S003 T039 new 0.44 NaN 0.26
S004 T038 new 0.67 NaN 0.03
S004 T019 new 0.78 NaN 0.08
S004 T031 new 0.62 NaN 0.08
S005 T038 new 0.76 NaN 0.06
S005 T014 new 0.60 NaN 0.10
S005 T018 new 0.84 NaN 0.14

Full tables: recommendations.csv, clusters.csv, summary.json.