CMCP-5 is an expert-guided human/machine hybrid “mind” applied to mental condition assessments. Employing the diagnoses of many psychiatric experts, and digitizing diagnostic approaches from multiple perspectives, CMCP-5 has effectively combined collective knowledge to determine a low-bias and highly-informed predisposition and probability assessment. The approach employs machine-learning to calibrate/correlate and correlate mental condition ground truths with a purpose-built Reference Database of individual personal “signatures” to determine which combinations of biometric variables most accurately predict mental condition predispositions and probabilities. New incoming records can be compared to this matrix to predict high-accuracy, evidence-based scores.
York University’s Clinical Study has generated these results (the following chart predicts depression probabilities using three different approaches):
The above Chart shows that the machine-learning results corroborate the expert-guided results and that the expert-guided approach outperforms the machine-learning approach. Both approaches used the same “ground truth” targets.
- The original YMI “expert-guided” approach where the predictive scoring is guided by psychoanalytical experts in advance of the application of machine-learning; and
- two “supervised” approaches where no expert guidance is provided in advance of machine-learning application.
| Model | YMI Expert guided | ML supervised | ML & DNA supervised |
| Statistical Significance (lower=better) | 0.021 | 0.050 | 0.045 |
| Accuracy | 84% | 81% | 84% |
| Precision | 86% | 85% | 90% |
| Recall | 98% | 75% | 77% |
| F1 Score (best overall) | 91% | 80% | 82% |