Prediction of malfunctioning time in computer systems

In this project, we present a process of a predictive maintenance algorithm designed to predict system failures based on system's historical data. This type of prediction has great importance preventing system downtime that can cost a lot in terms of time and money, and may be critical in cases of medical systems.
We designed LSTM models and an improved model using RUSboost algorithm. LSTM networks are a variation of RNN (recurrent neural networks) that are suited to time series data. RUSboost is an algorithm belongs to the boosting algorithms family, which is suitable for imbalanced data.
The results we achieved in this project are 94% in the recall metric and 83% in the F1 metric.

maintanance

Prediction of malfunctioning time in computer systems

Collaboration:

Refael logo