Water Quality Prediction using Machine Learning

Abstract

Freshwater is a critical resource for agriculture and industry's survival. Examination of water quality is a fundamental stage in the administration of freshwater assets. As indicated by the World Health Organization's yearly report, many individuals are getting sick or some are dead due to the lack of safe drinking water, especially pregnant ladies and kids. It is critical to test the quality of water prior to involving it for any reason, whether it is for animal watering, chemical spraying(Pesticides etc), or drinking water. Water quality testing is a strategy for finding clean drinking water. Accordingly, appropriate water monitoring is basic for safe, clean, and sterile water. Water testing is fundamental for looking at the legitimate working of water sources, testing the safety of drinking water, identifying disease outbreaks, and approving methodology and safeguard activities. Water quality is a proportion of a water’s readiness for a specific utilize in view of physical, chemical, and biological qualities.

Introduction

I. INTRODUCTION

Water is the principal source for shipping energy to each cell in the body and is additionally the regulator of all body capacities. The cerebrum contains 80% of water. Extreme drying out may prompt mental hindrances and loss of capacity to obviously think. Water is one of the most fundamental regular assets for the endurance of the whole life on this planet. In light of the nature of water, it tends to be utilized for various purposes like drinking, washing, or water system. Plants and creatures likewise rely upon water for their endurance. To put it plainly, all living organic entities need an enormous amount and great nature of water for presence. Freshwater is a fundamental asset to horticulture and industry for its essential presence. Water quality observation is a key stage in the administration of freshwater assets.  As indicated by the yearly report of WHO, many individuals are kicking the bucket because of the absence of unadulterated drinking water particularly pregnant ladies and youngsters. It is critical to check the nature of water for its expected reason, whether it be animals watering, compound showering, or drinking water.

Water quality testing is a device that can be utilized to find unadulterated drinking water. Consequently, the right checking of water is incredibly much significant for protecting unadulterated, and clean water. Water testing assumes a key part in breaking down the right activity of water supplies, testing the wellbeing of drinking water, perceiving sickness flare-ups, and approving cycles and precaution measures. Water quality is the proportion of the reasonableness of water for a specific reason in view of explicit physical, substance, and organic attributes.

Testing the nature of a water body, both surface water, and groundwater, can assist us with responding to inquiries concerning whether the water is satisfactory for drinking, washing, or water system to give some examples of applications. It can utilize the consequences of water quality tests to look at the nature of water starting with one water body and then onto the next in a locale, state, or across the entire country. Microbiological quality is for the most part the main pressing concern on the grounds that irresistible infections brought about by pathogenic microorganisms, infections, helminths, and so on are the most well-known and boundless wellbeing risk connected with drinking water. Overabundance amount of certain synthetic substances in drinking water prompts well-being risk. These synthetics incorporate fluoride, arsenic, and nitrate. Safe drinking (consumable) water should be passed on to the client for drinking, food game plan, individual neatness, and washing. The water ought to satisfy the normal quality rules for making it pure at the spot of supply to the clients.

One of the greatest machine (statistical) learning algorithms for pragmatic applications is Breiman's random forests (RF). Despite its practical usefulness, random forests remained fairly obscure until recently when compared to other AI and machine learning techniques, with little use in water research, particularly hydrological applications.

As a result, the power of ‘Breiman's’ unique calculation and its variations in water assets, and applications remains un-utilized. Aside from the standard applications of RF-based calculations in relapse and grouping issues, as well as the calculation of significant measurements, heir utilization for decile expectation, endurance investigation, and offhand surmising, appear to be less well known among water researchers and professionals. Random forest is accepted to have a place with the class of data driven models with regards to water resources.

Literature Survey shows a few current studies on the use of information-driven models in water assets, water asset design, and hydrology, where the random forest is lacking and a significant portion of the text is dedicated to brain organizations. Some of the most often discussed topics in literature on information-driven models are expectation (or determination), preprocessing, variable selection, splitting the dataset into preparation and testing phases, and predictive execution assessments.

II. LITERATURE SURVEY

III. RANDOM FOREST

Breiman's random forest (RF) notion, as well as related ideas and results, are covered in this section. To summarise, Breiman's RF calculation is distinguished from earlier RF executions by the use of characterization and relapse trees as base students; see Section 3.1 below. For convenience and without sacrificing unanimity, we will now accept the RF boundary documentation used in the randomForest R package, which is directly linked to Breiman's original work.

A. How Random Forest Works

A few works and reading materials, for example, Breiman, Biau, and Scornet, as well as James et al., Hastie et al., Kuhn and Johnson, have detailed introductions to RF calculations. A random forest is an AI computation that combines the concepts of classification and regression trees, as well as bagging and randomization. These principles are presented in Segment 3.1.1-Section 3.1.3, and Section 3.1.4 discusses how and why they are connected.

1. Supervised Learning

Supervised learning is one of the most generally practised parts of machine learning(ML) that utilizations named preparing information to assist models with making precise expectations. The preparation information here fills in as a boss and an educator for the machines, subsequently the name. A comparative system is instrumental in settling genuine difficulties, for example, picture classification, spam separating, risk evaluation, and so on. A supervised learning algo generally has an objective or result variable (or ward variable), which is recognized from a gave set of indicators (free factors). The calculation utilizes this arrangement of factors to make a capacity that guides contributions to wanted yields. This preparing system is repeated however long it takes for the model to accomplish an elevated degree of precision. Under the umbrella of supervised learning fall: Classification, Regression and Forecasting. Supervised learning frequently require non-trivial dataset sizes to advance dependably from ground truth perceptions. Models might require a large number of information and result guides to gain from to really perform. Bigger datasets, including more prominent quantities of noteworthy models from which to learn, empower the calculations to integrate an assortment of edge cases and produce models that handle these edge cases exquisitely. Contingent upon the business main concern, numerous long stretches of information are important to represent irregularity.

2. Classification and Regression Trees

In classification tasks, the AI program should reach a determination from observed values and decide to what classification new observations belong. For instance, while filtering messages as 'spam' or 'not spam’, the program should check out at existing observational information and channel the messages appropriately. In regression tasks, the AI program should assess - and comprehend - the connections among variables. Regression analysis centers around one ward variable and a progression of other changing factors - making it especially valuable for expectation and determination.

Fig. 1:- Decision tree example. “Xj denotes predictor variables. The tree has four internal nodes and five leaves (terminal nodes). Xj ≤ tk and Xj > tk correspond to the left and right branches of each internal split, respectively. Ri denotes the mean of the observations at leaf I”.

The decision criteria for node portions in regression trees are tuned/advanced by streamlining the number of squared deviations, while the Gini record is streamlined in order. It's worth noting that, on the whole, tree-based computations (counting CARTs) are quite raucous, with big differences in the splitting rules and the tree's size having been identified.

3. Bagging

Bagging (truncation for bootstrap aggregation) is a Breiman-proposed outfit learning approach. It generates a bootstrap test from the initial data and then uses the test to generate a model (e.g., a CART). The system is ntree times rehashed. It is the normal to Bag's forecast of the ntree prepared models' expectations. Sacking reduces the discrepancy between expected and actual effort in this way, but it still requires unprejudiced models to function well.

4. Random Forests

Random Forest is bagging of CARTs with some extra level of randomization. Stowing of

Trucks are expected to reduce their precariousness (see Section 3.1.2). Further, randomization is utilized to decrease the relationship between the trees and, thusly, lessen the fluctuation of the expectations (i.e., the normal of the trees). Randomization is led by haphazardly choosing to try indicator factors as a contender for parting. Forecast in relapse is performed by averaging the expectations of each tree, while in the order it is performed by acquiring the larger part class vote from the singular tree class votes. A possibility for boundary tuning of arbitrary backwoods is to use out-of-sack (OOB) blunders. Out-of-pack tests (around 1/3 of the preparation set, see Biau and Scornet) are the examples staying in the wake of bootstrapping the preparation set. The previously mentioned strategy looks like the notable k-overlay cross-approval.

III. DECISION TREE

Decision tree is the essential design of numerous group learning techniques, called order tree when utilized for characterization and relapse tree when utilized for relapse. In contrast to other order strategies that consolidate a bunch of highlights in a solitary choice advance to perform grouping, choice trees depend on a multi-stage or various leveled choice plan or tree construction and comprise of nodes and coordinated edges. Decision tree comprises of two sorts of nodes: inner nodes and leaf nodes. Component and trait information addressed with “inside node” and the “leaf node” addresses a class. In particular, a “leaf node” address consequence of a choice from the root node to this leaf node, and an inside node addresses the information characterization test performed by then, i.e., the element or on the other hand trait to which the test information has a place. Every node of the decision tree structure makes a paired choice, and the examples contained in the node in view of the consequence of the quality test are separated into sub-node (the root node contains the full arrangement of tests), and the way from the root node to each leaf node relates to a grouping of choice tests. This handling is generally performed by dropping down the tree until a leaf node is reached. In the choice tree approach, the attributes of the information (i.e., other water boundaries) are indicator factors, while the class to be planned (TN) is alluded to as the objective variable. The tree-like construction of a decision tree is displayed in Fig 2.

There are basically 10 steps for making our model predict the water quality of the water samples. Those steps are:-

A. Problem Identification

In this step, we identify the problem which is solved by our model. So the problem to be solved by our model is water quality prediction using a dataset.

B. Data Extraction:-

In this, we extract the data from the internet to train our data and predict the water quality. So for that, we take the CPCB(Central Pollution Control Board India) dataset which contains 3277 instances of 13 different wellsprings which are collected between 2014 to 2020.

C. Data Exploration:-

In this step, we analyze the data visually by comparing some parameters of water with the WHO standards of water. It gives a slight overview of the data.

D. Data Cleaning

In this step, we clean that data like if there are some missing values in it so we replace them with mean and remove noise from the data.

E. Data Engineering

In this step, we ensure that the data is quality data so that the prediction accuracy increases.

F. Data Selection

In this step, we select the data types and source of the data. The essential goal of data selection is deciding fitting data type, source, and instrument that permit agents to respond to explore questions sufficiently

G. Data Splitting

In this step, we divide the dataset into smaller subsets for easing the complexity. Normally, with a two-section split, one section is utilized to assess or test the information and the other to prepare the model.

H. Data Modeling

In this step, we create a graph of the dataset for visual representation of data for better understanding. A Data Model is this theoretical model that permits the further structure of conceptual models and to set connections between data.

I. Model Evaluation

Model Evaluation is a fundamental piece of the model improvement process. In this step, we evaluate our model and check how well our model do in the future.

VI. PARAMETERS

Parameters are the basic and most important component of a model. Based on the parameter model prediction and it shows the skill of the model over the data. Similarly, random forest algorithms also have some parameters to predict water quality. We take 10 parameter to predict the water quality. Those parameters are :-

1. Ph
2. Solid
3. Cholride
4. Conductance
5. Sulphate
6. Hardness as CaCO3
7. Trihalomethanes
8. Turbidity
9. Organic Carbon
10. Potability

To predict whether the water is drinkable or not WHO provides some standard values for these parameters of water which is as follows,

Table:- WHO standard for parameters

VII. WATER QUALITY INDEX

WQI is the correlation of the sum with an erratic or logical norm or with a pre-determined base. In this way, the WQI observed and announced natural status and patterns on guidelines quantitatively. A water quality list is a way to sum up a lot of water quality information into straightforward terms (e.g., great) for answering to the board and the general population in a predictable way. Notwithstanding the nonattendance of a universally acknowledged composite file of water quality, a few nations have utilized and are involving collected water quality information in the improvement of water quality lists.

To calculate the water quality index(WQI) conventionally we take 10 features of water to reflect the quality of water like ph, chloride, conductance, etc. In this paper, we use all 10 parameters to calculate the WQI of the water. The general formula to calculate the water quality index is given below,

Conclusion

As we all know the importance of water for the human body. So knowing the Quality of the water is very much necessary because if we drink water without knowing that it is safe for drinking we could get sick. There are plenty of water-borne diseases like Cholera, Typhoid, Giardia, E. Coli, Hepatitis A, and so on. These types of diseases happen if we drink non-drinkable water. So knowing the quality of the water is the most important thing. But the main problem lies here. For testing the water quality we have to conduct lab tests on the water which is costly and time-consuming as well. So, in this paper, we propose an alternative approach using artificial intelligence to predict water quality. This method uses a significant and easily available water quality index which is set by the WHO(World Health Organisation). The data taken in this paper is taken from the PCPB India which includes 3277 examples of the distinct wellspring. In this paper, WQI(Water Quality Index) is calculated using AI techniques. So in future work, we can integrate this with IoT based framework to study large datasets and to expand our study to a larger scale. By using that it can predict the water quality fast and more accurately than any other IoT framework. That IoT framework system uses some limits for the sensor to check the parameters like ph, Temperature, Turbidity, and so on. And further after reading this parameter pass these readings to the Arduino microcontroller and ZigBee handset for further prediction. This proposed IoT system would recognize the correct quality of water and raise some alarm to the concerned professional for further action. It prevents people from drinking undrinkable water and reduces the danger of having water-borne diseases like typhoid to happen. So by this way, we can use this model for future workshops on water quality and raise awareness

References

[1] S. Geetha, S. Gouthami. Internet of Things Enabled Real Time Water Quality Monitoring System ,2017. [2] Umair Ahmed, Rafia Mumtaz, Hirra Anwar, Asad A. Shah, Rabia Irfan and José García-Nieto. Efficient Water Quality Prediction Using Supervised Machine Learning, 2019. [3] Ashwini K, D. Diviya, J.Janice Vedha, M. Deva Priya.Intelligent Model For Predicting Water Quality. [4] A.N.Prasad, K. A. Mamun, F. R. Islam, H. Haqva.Smart Water Quality Monitoring System, 2015 [5] Hadi Mohammed, Ibrahim A. Hameed, Razak Seidu.Machine Learning: Based Detection of Water Contamination in Water Distribution systems,2018 [6] Priya Singh,Pankaj Deep Kaur.Review on Data Mining Techniques for Prediction of Water Quality,2017 [7] Manish Kumar Jha,Rajni Kumari Sah,M.S.Rashmitha,Rupam Sinha,B.Sujatha.Smart Water Monitoring System for Real-Time Water Quality and Usage Monitoring,2018 [8] Water QualityMonitoring System using IoT and Machine Learning, in Proceedings of the IEEE International Conference onResearch in Intelligent and Computing in Engineering, pp.1-5, 2018 [9] Amir Hamzeh Haghiabi, Ali Heidar Nasrolahi, Abbas Parsaie. Water quality prediction using machine learning methods, 2018 [10] A Gollapalli, Mohammed; Ensemble Machine Learning Model to Predict the Waterborne Syndrome, 2022. [11] Breiman, L. Random forests. Mach. Learn. 2001 [12] Biau, G.Ã.Š.; Scornet, E. A random forest guided tour. TEST 2016, 2 [13] Liakos, K.; Busato, P.; Moshou, D.; Pearson, S.; Bochtis, D. Machine learning in agriculture: A Review. Sensors 2018 [14] Gislason, P.O.; Benediktsson, J.A.; Sveinsson, J.R. Random forests for land cover classification. Pattern Recognit. Lett. 2006 [15] Maxwell, A.E.; Warner, T.A.; Fang, F. Implementation of machine-learning classification in remote sensing: An applied review. Int. J. Remote Sens. 2018 [16] Chen, X.; Ishwaran, H. Random forests for genomic data analysis. Genomics 2012 [17] Boulesteix, A.L.; Janitza, S.; Kruppa, J.; König, I.R. Overview of random forest methodology and practical guidance with emphasis on computational biology and bioinformatics. Wiley Interdiscip. Rev. Data Min. Knowl. Discov. 2012 [18] Solomatine, D.P.; Ostfeld, A. Data-driven modelling: Some past experiences and new approaches. J. Hydroinformatics 2008 [19] Gazzaz, N.M.; Yusoff, M.K.; Aris, A.Z.; Juahir, H.; Ramli, M.F. Artificial neural network modeling of the water quality index for Kinta River (Malaysia) using water quality variables as predictors. Mar. Pollut. Bull. 2012 [20] Sakizadeh, M. Artificial intelligence for the prediction of water quality index in groundwater systems. Model. Earth Syst. Environ. 2016 [21] Abyaneh, H.Z. Evaluation of multivariate linear regression and artificial neural networks in prediction of water quality parameters. J. Environ. Health Sci. Eng. 2014 [22] Kangabam, R.D.; Bhoominathan, S.D.; Kanagaraj, S.; Govindaraju, M. Development of a water quality index (WQI) for the Loktak Lake in India. Appl. Water Sci. 2017 [23] Srivastava, G.; Kumar, P. Water quality index with missing parameters. Int. J. Res. Eng. Technol. 2013, [24] Jayalakshmi, T.; Santhakumaran, A. Statistical normalization and back propagation for classification. Int. J. Comput. Theory Eng. 2011 [25] Liaw, A.; Wiener, M. Classification and regression by randomForest. R News 2002 [26] Günther, F.; Fritsch, S. Neuralnet: Training of neural networks. R J. 2010 [27] Singh, J.; Yadav, P.; Pal, A.K.; Mishra, V. Water pollutants: Origin and status. In Sensors in Water Pollutants Monitoring: Role of Material; Springer: Berlin/Heidelberg, Germany, 2020 [28] Jiang, J.; Tang, S.; Han, D.; Fu, G.; Solomatine, D.; Zheng, Y. A comprehensive review on the design and optimization of surface water quality monitoring networks. Environ. Model. Softw. 2020 [29] Park, J.; Kim, K.T.; Lee, W.H. Recent advances in information and communications technology (ICT) and sensor technology for monitoring water quality. Water 2020

Copyright

Copyright © 2022 Nishant Rawat, Mangani Daudi Kazembe, Pradeep Kumar Mishra. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.