A Machine Learning Approach towards Automatic Water Quality Monitoring

Abstract

Increasing rate of water pollution and consequently waterborne diseases are the engrossing evidence towards danger to living organisms. It becomes a great challenge these days to preserve our flora and fauna by controlling various unexpected pollution activities. Although the invention of many schemes and programmes regarding water purification has done a tremendous job, but still there is something that has been lagging. With increase in population, industrialization and global warming situation is getting worse day by day. It becomes very difficult to get safe drinking water and appropriate quality water for other domestic usage and agriculture purpose. Major reasons for water pollution include undesirable increase in impurities. These may cause eutrophication of the water body, change in taste, discolouration and odour of water, water borne diseases, increase in water toxic nature etc. For water to be serviceable it should be aesthetically acceptable, chemically safe, bacteria free; organic substances and radioactive elements should be absent. So, there is an urgent need to look into this situation and take the corrective and necessary actions to overcome this situation. The government is paying an attention to this problem and finding the ways to control the situation. However, major areas are not developed to the point and water quality estimation is totally dependent upon sampling at location and testing in laboratories. Manual sampling and measurements are prone to human errors and these techniques may create ambiguities in predicted output. In this paper we have presented Machine Learning (ML) approach for calculating the Water Quality Index (WQI) and classification of water quality to estimate water characteristics for usage. For analysis, decision tree method is used to estimate water quality information. The standard values of parameters are selected as per guidelines provided by World Health organization (WHO). Results calculated using ML techniques showed prominent accuracy over traditional methods. Accuracy achieved is also significant, i.e. 98 %. Likewise, projection of gathered data was done utilizing web interface and web app to alert the authorities about contamination.