Algorithm 1 PGraphDD-QM: concept drift detection using quality metrics

Require:  Event stream: $\mathcal{S}$, LSTM model: $\mu$, Threshold: $\varphi$ 1:$fScoreLog\leftarrow \left[\right]$, $DriftLog\leftarrow \left[\right]$ 2:    {read event stream, split windows, prepare data, train LSTM model and make predictions} 3: $e\leftarrow fetch\left(eventStream\right)$ 4:$w\leftarrow WinSize$ {the window size} 5: $parts\leftarrow splitLog(e,w)$ 6:for alli in range(0, parts) do 7:   $j\leftarrow i+1$ 8:   if $i=parts-1$ then 9:     $j\leftarrow 0$ 10:    $\mathcal{R}\leftarrow part\left(i\right)$, $\mathcal{D}\leftarrow part\left(j\right)$ 11:  end if 12:  while $w\ne 0$ do 13:    $prepare(\mathcal{R},\mathcal{D})$ 14:    ${\mu }^{\prime }\leftarrow (\mu ,\mathcal{R})$    {train an LSTM model on the reference window} 15:       {Apply the trained model to both reference and detection windows} 16:    $R_{pred}\leftarrow predict({\mu }^{\prime },\mathcal{R})$ 17:    $D_{pred}\leftarrow predict({\mu }^{\prime },\mathcal{D})$ 18:        {build next event prediction probability matrices, construct graphs} 19:    $P_{ij}^R\leftarrow generate(Pred\_Prob\_Mat,R_{pred})$ 20:    $P_{ij}^D\leftarrow generate(Pred\_Prob\_Mat,D_{pred})$ 21:    $G^R\leftarrow drawGraph(i,o,N,E,P_{ij}^R,R_{pred})$ 22:    $G^D\leftarrow drawGraph(i,o,N,E,P_{ij}^D,D_{pred})$ 23:        {calculate the quality metrics} 24:    $p\leftarrow performance\left(\right)$ 25:    $fitness\leftarrow p.fitness\left(\right)$ 26:    $precision\leftarrow p.precision\left(\right)$ 27:    $fScore\leftarrow p.fScore(fitness,precision)$ 28:    if $fScore\le \varphi$ then 29:       $print(“driftfoundinwindo{w}^{”}+str\left(j\right))$ 30:       $Report\leftarrow (i,j,fScore)$ 31:    end if 32:  end while 33: end for