An IoT based efficient Air pollution prediction system using DLMNN classifier

The Internet of Things (IoT) is a global system of "smart devices " which senses and connects with their environment and communicates with users as well as other systems. Air Pollution (AP) is one of the most significant global issues. Prevailing AP systems have low accuracy and require laboratory-based analysis. Therefore, improved prediction systems are needed. To overcome such problems, this paper proposes an IoT based efficient APprediction system utilizing the Deep Learning Modified Neural Network (DLMNN) classifier. Initially, the faulty node detection is done in the sensor nodes using the H-ANFIS algorithm. Here, ANFIS is hybridized with the K-Medoid algorithm. After that, the features are extracted from the sensed data and the unnecessary features are reduced by using the MPCA algorithm. Next, based on the reduced features, the data are balanced by using Entropy-HOA. Then, the balanced sensed data are pre-processed using replacing of missing attributes and HDFS. Next, the pre-processed data are tested with an AP prediction system employing the DLMNN classifier, where the Pity Beetle Algorithm (PBA) is used for weight optimization. Then, the predicted result is stored in the cloud server. Finally, the stored data is visualized. Experimental results have proved that the proposed system gives a better result than the existing systems.