Search for a command to run...
This study aimed to analyze the quality of water and to predict water quality index and water quality class by using AI algorithm (ANN). To overcome the task, the Artificial Neural Network was used to model and predict four water quality metrics, including pollution level. These variables are, in order of importance, pH, temperature, DO, electrical conductivity (EC), Fluoride (F) and total solids (TS). Data are analyzed step by step, first to collect data to ensure that data is consistent with time intervals, clean data by removing and handling unnecessary and missing values and to normalize and standardize the data. The traditional method of estimating water quality involves costly and time-consuming statistical and laboratory analyses. Due to the concerning effects of low water quality, a speedier and less expensive alternative approach is required. For this reason, this study investigates a number of supervised machine learning algorithms to estimate the water quality class (WQC), a unique class defined based on the WQI, and the water quality index (WQI), a singular index to describe the overall quality of water. Four input factors are used in the suggested methodology: pH, Temp, DO, EC, Fluoride, and TS. The pH values, which range from 7 to 8.4, were found to be within the standards for water quality. However, the parametric values for temperature (°C), DO (mg/l), EC (dS/m), Fluoride and TS (mg/l) (mg/l) ranged from 18 to 88, 5.5 to 8.5, 80 to 840, 0.01 to 2.3 and 680 to 8580, respectively. Overall, the artificial neural network did a pretty good job of modelling and predicting the real water quality data set. The training model performance evaluation shows that the R2 values for pH, Temp, DO, EC, Fluoride, and TS are 0.7467, 0.6682, 0.7395, 0.4425, 0.3246, and 0.5525. The results of the testing model performance indicate that the R2 values for pH, Temp, DO, EC, Fluoride, and TS are 0.9984, 0.943, 0.7954, 0.3817, and 0.8313, respectively, while the results of the forecast performance evaluation indicate that the R2 values for these same parameters are 0.0837, 0.953, 0.983, 0.41, 0.702, and 0.64. It was observed that the Root Mean Squared Error (RMSE) values for pH, temperature, DO, EC, Fluoride, and TS were 0.055, 0.099, 0.05, 0.049, 0.077, and 0.068. The study findings will help researchers working on the water quality index. Some assessment indicators were calculated to evaluate the regression models’ effectiveness: The coefficient of determination (R2), mean absolute error (MAE), Relative absolute error (RAE), mean square error (MSE), Mean Percentage Error (MPE), Mean Absolute Percentage Error (MAPE), Relative root Mean Square Error (RRMSE), Relative Squared Error (RSE), Mean Square Percentage Error (MSPE) and Mean Absolute Error (MAE). For a variety of reasons, determining water suitability requires predicting Water Quality Index (WQI) and Water Quality Classification (WQC) using machine learning models for Water quality monitoring at the right instant as compared to traditional laboratory analysis, predictive models provide real-time or nearly real-time estimation of WQI and WQC, which is more efficient and economical. To determine how effective the created model was, performance evaluation of the machine learning models. Evaluation parameters based on confusion matrices were taken into consideration in order to assess the performance of the discussed models. We employed precision and accuracy in the classification of water quality. This study utilized Orange—machine learning technique and four prediction models, including linear regression, tree, KNN, and SVM and suggested an intelligent real-time water quality monitoring strategy and concentrated on quantifying and classifying water quality using machine learning techniques. The obtained WQI value was 50.83. This WQI score indicates that the water quality class (WQC) is lies in medium category, according to the acquired WQI score. The Water Quality Index (WQI) score and Water Quality Class (WQC) were quite good. The algorithm technique used for calculating water quality index was very helpful in determining the concentrations of all parameters, even if we have missing values.