Efficient IOT based Water Quality Prediction Using Cat Swarm Optimized Neural Network classification

Water is the most significant sources for human life, but, it is in serious threat of contamination by life itself. The protection and availability of drinking-water are major worries throughout the globe. In this work,anIOT based solution isintroduced to check and predict the water quality and alert the user before the water gets polluted. The proposed system uses IoT and optimized neural network for prediction. It consists of various embedded sensors like conductivity, pH, turbidity and color. The measured sensor values are stored in the database and further directed for prediction analysis. The Cat swarm optimization (CSO) based neural network algorithm is used for forecasting thequality result. The proposed system alerts the user when any of themeasured parameters are lesser than the fixed thresholds. This technique can also be implemented in water plants, rivers and industries. Article Received: 20 September 2020, Revised: 30 November 2020, Accepted: 18 December 2020


Introduction
Currently, water quality has become more severe threats. The quality expectation of water, as a basic part of the water environment controlling, is to find the consistency for the determined file with the time utilizing certain guaging approaches, and to understanding the development style of the water quality dependent on the past information. Presently, the estimating procedures normally utilized in water quality contains dim framework, exponential smoothing strategy, the neural organization, the numerous straight relapse, the numerical expectation model of the water quality and so forth.
The explanation is that there is an assortment of dubious impact factors inside the water condition structure. The Back Propagation Neural Network has been regularly applied as a pragmatic nonlinear incredible framework anticipating and showing mechanical get together, regardless, there are also a couple of distortions. For example, the organization structure is hard to pick, and it is essential for the organization sort out how to trap into a neighborhood least. Moreover, the learning step should be gone after for set and alter, and it is skewed to happen for a moderate social occasion or a non-mix.
In this work, an IOT based system is proposed to check and predict the water quality and alert the user earlierto avoid water contamination. (2)Randomly create N felines (plan sets) with arbitrary speed .The speed ought to be littler than a fixed most prominent speed esteem.
(3)Randomly portray the felines into looking for and following modes .From all out populace, 20% felines considered following mode and 80% felines took into consideration looking for mode.
(4)Evaluate the fitness function for desired response .
(5)The cats at that point move to either seeking or tracing mode.
(6) Check final solution obtained. If desired output reached terminate the iteration otherwise repeat step 2 to 5.
In any case, objective of cat swarm is defined by : , where 0 < i < j If the objective is minimization, then FSb = FSmax; otherwise, FSb = FSmin.

3.
Proposed system In the proposed system integrates various embedded sensors such as pH sensor, coloursensor, Turbidity sensor, conductivity sensor and DO sensor etc. A Wi-Ficomponent is used to transfer data from arduino to cloud server. The sensors unceasingly transfer the data to the controller.The collected data is processed by optimized neural classifier and alert user to avoid water contamination earlier.

Hardware design
Controller An Arduino UNO is used as a core one. The Arduino victimized here is mega 328 because of multiple analog pins. Sensors for monitoring pH sensor -→ used to monitor pH levels it measures the hydrogen-ion density in a bleach Temperature sensor --) LM35 for measuring temperature Wi-Fi module → wireless meshwork that is connected to the Internet IoT platform and Neural network models →The quality boundaries are marked datasets with wanted yields of a particular blend of information sources. The neural organization will create yield to arrange water quality as risky, be cautious, and acceptable. Proposed CSO-NN There are numerous methods for training a neural network. By far the most general technique is called backpropagation. Though mathematically elegant, backpropagation is not suitable. This work designates an alternative neural network training technique that uses CWO. In proposed CSO optimized NN, weights and bias of both hidden and output layer updated using CSO algorithm. And error is displayed before and after optimization by CSO. Implementation results

Conclusion
The work presents the low cost and accurate water quality prediction system to avoid pollution of water. The contamination status of water is examined using sensorbased IoT system and the future forecasting of water pollution is obtained using optimized neural classification. The proposed IOT solution comprises of sensors connected to the IoT server to analyze the condition of water. And the warning data is sent to the client before the water gets over pollution. The proposed predicting and alerting system supports to save the water from pollution and is also inexpensive