Support Vector Machine Classifier of Grain Stored in Silo-bags by Using Wireless Network of Temperature and Moisture Sensors

We propose using support vector machines (SVMs) as a machine learning classifier in a grain-conditionmonitoring system. This system is used to proactively monitor the silobag-stored grain condition. The system consists of a wireless sensor network (WSN) and a machine learning classifier. As far as the hardware is concerned, the system relies upon a wireless network of temperature and moisture sensors; and as far as the software is concerned, SVM is used to classify the grain status according to the data coming from the WSN. The grain condition is classified into three classes: safe, risky and dangerous. We use two different multiclass classification algorithms together with SVM. The first algorithm is the one-againstall algorithm and the second is the error-correcting output codes (ECOC). Both algorithms accurately classify the grain status such that the proportion of the misclassified cases from the testing data is zero and, therefore, the classification accuracy is 100%. Nevertheless, the ECOC algorithm is faster than the one-against-all algorithm in terms of the training and testing time. In addition, we compare our results with a previous work, which used artificial neural networks (ANNs) as the classifier, and the current results are obviously better than the previous results.