Table 4 Performance metrics for model evaluation
From: Nutritional management recommendation systems in polycystic ovary syndrome: a systematic review
Metric | Formula | Range | Desirable |
|---|---|---|---|
𝑃𝑟𝑒𝑐𝑖𝑠𝑖𝑜n | 𝑇𝑃/ 𝑇𝑃 +𝐹𝑃 | 0-100% | Max |
Sensitivity/ Recall | 𝑇𝑃 /𝑇𝑃 + 𝐹𝑁 | 0-100% | Max |
Specificity | 𝑇𝑁 /𝑇𝑁 + 𝐹𝑃 | 0-100% | Max |
Accuracy | 𝑇𝑃 + 𝑇𝑁/𝑇𝑃 + 𝑇𝑁 + 𝐹𝑃 + 𝐹𝑁 | 0-100% | Max |
𝐹 − 𝑀𝑒𝑎𝑠𝑢𝑟e | (2 * Precision * Recall) / (Precision + Recall) | 0–1 | Max |
Positive Predictive Value (PPV) | (sensitivity * prevalence) / [ (sensitivity * prevalence) + ((1 – specificity) * (1 – prevalence)) | 0-100% | Max |
Kappa statistic | 2*(𝑇𝑃 *𝑇𝑁- 𝐹𝑁* 𝐹𝑃)/(𝑇𝑃+ 𝐹𝑃)*(𝐹𝑃+ 𝑇𝑁)*(𝑇𝑃+ 𝐹𝑁)*(𝐹𝑁+ 𝑇𝑁) | 0–1 | Max |
Mean Absolute Error (MAE) | (1/n) Σ(i = 1 to n) |y_i – ŷ_i| | LB:0 | Min |
Root Mean Squared Error (RMSE) | √ Σ(i = 1 to n) (y_i – ŷ_i)2/𝑁 | LB:0 | Min |
Root Relative Squared Error (RRSE) | √Σ(i = 1 to n) (ti − ri)2 /Σ(i = 1 to n) (ti−ˉt)2 | 0-∞ | Min |
Area Under the Curve (AUC) | 0–1 | Max | |
Precision-Recall Curve (AUPR) | 0–1 | Max |