Multi-Stage Feature Selection Based Intelligent Classifier for Classification of Incipient Stage Fire in Building

Abstract

In this study, an early fire detection algorithm has been proposed based on low cost array sensing system, utilising off- the shelf gas sensors, dust particles and ambient sensors such as temperature and humidity sensor. The odour or “smellprint” emanated from various fire sources and building construction materials at early stage are measured. For this purpose, odour profile data from five common fire sources and three common building construction materials were used to develop the classification model. Normalised feature extractions of the smell print data were performed before subjected to prediction classifier. These features represent the odour signals in the time domain. The obtained features undergo the proposed multi-stage feature selection technique and lastly, further reduced by Principal Component Analysis (PCA), a dimension reduction technique. The hybrid PCA-PNN based approach has been applied on different datasets from in-house developed system and the portable electronic nose unit. Experimental classification results show that the dimension reduction process performed by PCA has improved the classification accuracy and provided high reliability, regardless of ambient temperature and humidity variation, baseline sensor drift, the different gas concentration level and exposure towards different heating temperature range.

Multi-Stage Feature Selection (MSFS) Algorithm for UWB-Based Early Breast Cancer Size Prediction

“…Based on the comprehensive review done on the previous researches, five data normalization methods are chosen from the commonly used methods, namely, Decimal Scaling (DS), Z-score (ZS), Linear Scaling (LS), Min-Max (MM) and Mean & Standard Deviation (MSD) methods [19][20][21][22]. The other five data normalization methods are newly introduced in early breast cancer detection application, namely, Relative Logarithmic Sum Squared Voltage (RLSSV), Relative Logarithmic Voltage (RLV), Relative Voltage (RV), Fractional Voltage Change (FVC) and Relative Sum Squared Voltage (RSSV) [8][9]. These data normalization methods are proposed by [8][9] to overcome the baseline drift error that normally comes together with the data sample which affects the quality of the data samples.…”

“…The highest peak of the signal is approximately at 4.3 GHz same as the center frequency of the UWB antenna used. It is divided into multiple stages [8]. Once the data is preprocessed, it is normalized to 10 different data normalization methods, and the data is reduced using PCA.…”

“…The exploration and exploitation of the data will be insufficient during the feature selection as the features are reduced at the initial stage. As a result, only some redundant features are selected, and some useful features are lost due to poor data management [8][9].…”

“…The proposed Multi-Stage Feature Selection (MSFS) can be a solution in overcoming the mentioned drawbacks of single-stage feature selection method. MSFS can increase the learning model performance in breast cancer detection application [8][9]. The proposed multi-stage feature selection method will be discussed comprehensively throughout the paper.…”

Pollutant Recognition Based on Supervised Machine Learning for Indoor Air Quality Monitoring Systems

“…Baseline manipulation is the solution to the problem and the correct way of representing the signal when the analysis deals with sensor values from different conversion units. Baseline manipulation helps to pre-process the sensor output to free itself from the drift effect, the intensity dependence and, possibly, from non-linearity [40,41]. The details about other type of data pre-processing techniques will be described in the next section.…”