scala - MLlib: Calculating Precision and Recall for multiple threshold values -

i set setting threshold value of logistic regression 0.5 before use scoring. want precision, recall, f1 score value. unfortunately, when try doing threshold values see 1.0 , 0.0. how metrics threshold values other 0 , 1.

for example here o/p:

threshold is: 1.0, precision is: 0.85

threshold is: 0.0, precision is: 0.312641

i don't precision threshold 0.5. here relevant code.

// setting threshold value of logistic regression model here.

model.setthreshold(0.5)  // compute score , generate rdd prediction , label values.   val predictionandlabels = data.map {    case labeledpoint(label, features) => (model.predict(features), label) } 

// want compute precision , recall , other metrics. since have set model threshold 0.5, want pr @ value.

val metrics = new binaryclassificationmetrics(predictionandlabels) val precision = metrics.precisionbythreshold()  precision.foreach {    case (t, p) => {     println(s"threshold is: $t, precision is: $p")      if (t == 0.5) {       println(s"desired: threshold is: $t, precision is: $p")             } } 

the precisionbythreshold() method trying different thresholds , giving corresponding precision values. since thresholded data, have 0s , 1s.

let's have: [0 0 0 1 1 1] after thresholding , real labels [f f f f t t].

then thresholding 0 have [t t t t t t] gives 4 false positive , 2 true positive hence precision of 2 / (2 + 4) = 1/3

now thresholding 1 have [f f f t t t] , gives 1 false positive , 2 true positive hence precision of 2 /(2 + 1) = 2/3

you can see using threshold of .5 give [f f f t t t], same thresholding 1, precision threshold 1 looking for.

this bit confusing because thresholded predictions. if not threshold predictions, , let's had [.3 .4 .4 .6 .8 .9] (to stay consistent [0 0 0 1 1 1] have been using).

then precisionbythreshold() give precisions values threshold 0, .3, .4, .6 .8 .9, because these threshold giving different results , different precisions, , value threshold .5 still take value next larger threshold (.6) because again, give same predictions hence same precision.