Recurrent Trend Predictive Neural Network for Multi-Sensor Fire Detection

We propose a Recurrent Trend Predictive Neural Network (rTPNN) for multi-sensor fire detection based on the trend as well level prediction and fusion of sensor readings. The rTPNN model significantly differs from existing methods due to recurrent data processing employed in its architecture. performs time series each reading captures trends multivariate produced by detector. compare performance with that Linear Regression (LR), Nonlinear Perceptron (NP), Multi-Layer (MLP), Kendall- $\tau $ combined MLP, Probabilistic Bayesian (PBNN), Long-Short Term Memory (LSTM), Support Vector Machine (SVM) publicly available set. Our results show outperforms all other models (with 96% accuracy) while it is only achieves high True Positive Negative rates (both above 92%) at same time. also triggers an alarm 11 s start fire, where this duration 22 second-best model. Moreover, we present execution acceptable real-time applications.