Search

In this study, it is predicted that there may be a certain order in the consumption of an important need such as drinking water by the household, as well as irregular consumption depending on different factors. Increasing population, limited drinking water resources, developing infrastructure and technology have increased the demand for drinking and utility water. There is a search for alternative water sources to meet this demand, but it is foreseen that these demands can be met by not wasting existing water and using it more efficiently. By using machine learning (ML) methods, which is a sub-branch of artificial intelligence (AI), drinking water consumption data in the past periods were analyzed, and ordinary and unusual consumption behavior models were extracted. It is envisaged that by detecting abnormal consumptions that may occur in drinking water consumption and informing the subscribers about this issue, it will be ensured that the consumption in the household remains within the normal consumption range. Although the amount of data collected, recorded and processed in today's IT world has increased significantly, it is known that the exact analysis is difficult in terms of time and cost. In this study, subscriber, meter, consumption, bill and payment data of 8,224 residential subscribers, whose water meter index reading is more than 160 periods throughout the province of Kayseri, between 2006 and 2022 (first 6 months) were taken into account. The data are combined on a spatial subscriber basis and a 41-features dataset is obtained. The dataset was transformed into a dataset with 24 features as a result of data preprocessing. In the study, 6 sub-datasets were obtained by using information gain (IG), gain ratio (GR), symmetric uncertainty coefficient (SU), pearson correlation coefficient (r), f-score and random forest (RF) feature selection methods. The 7th sub-dataset was obtained from the intersections of the selected features in the sub-datasets. In all datasets, abnormal and normal drinking water consumptions were determined by using 7 different ML anomaly analysis methods: tukey outlier labeling (TOL), forest of isolation (IF), z-score, copula-based outlier detection (COPOD), median absolute deviation (MAD), local outlier factor (LOF), and elliptical envelope (EE). At the beginning of the study were unsupervised drinking water consumption data at the end of the study, labeled as 4 different classes and the dataset was made supervised. Using the finally obtained supervised dataset, decision trees (DT), gaussian naive bayes (NB), k-nearest neighbors (KNN), logistic regression (LJR), multilayer perceptron neural network (MLP-NN), RF and gradient boosting (GB) have been developed consumption class estimation models with 7 different ML methods. As a result of the study, it has been proven that abnormal drinking water consumption can be detected by ML methods, and it has been revealed that necessary policies can be created for more efficient use of water without wasting water and measures can be taken for this. [ABSTRACT FROM AUTHOR]

In this study, the criteria of choosing a neighboring station by k-nearest neighbor method were evaluated through outlier detection. Monthly average temperature, average relative humidity and total precipitation data were used between 1926 and 2012 at 129 stations in Turkey. The first problem when choosing a neighbor according to the extent of nearness is that 0.04% - 3% of neighbors are more than 140 km away and less than 0.4 correlation because of choosing distant neighbors due to lack of data. The second problem is the more distant neighbors show a significantly higher correlation than the 5 closest neighbors at 0.1 - 3% of monthly series. When the neighbor is selected according to the correlation coefficient, the 1st neighbor of each month can be used for only that month in 2% - 8% of the stations. These percentages decrease for 2nd and 3rd neighbors. In 25% - 86% of stations, the 1st neighbors in certain months can be used for all months, but it should be determined which certain months are they. In addition, in 2.5% of the stations, the 1st neighbor of each variable can be used for only that variable. The percentage decreases for 2nd and 3rd neighbors. In 29% of the stations, the 1st neighbors in certain variables can be used for all the variables, but it should be determined which certain variables are they. Other than these, when the detected outliers are removed, the correlation coefficient becomes mostly 1-36% higher, sometimes 1-5% less than for the series with outliers. [ABSTRACT FROM AUTHOR]

In the multiple linear regression model, commonly encountered problems are multicollinearity and outlier problems. There are many studies about robust biased estimators in the literature to solve these problems simultaneously. The most commonly used of these estimators are robust Ridge estimators. Many studies are still carried out on the Ridge estimator, which depends on the biasing parameter. The performances of the previously proposed biased parameters were compared in the classical Ridge estimator. In this study, we have compared the performances of the biasing parameters for the robust Ridge GM estimators based on the simulation and real data studies. [ABSTRACT FROM AUTHOR]

With the introduction of computers into our lives, digital data sizes are increasing gradually. Non-standard values (outliers) which behave differently from the others can be found in these data produced in the digital world. Detection of these values, especially in big data sets; has great importance in fields such as security, insurance, finance, medicine and genetics. Clustering methods of data mining techniques are frequently used in outlier detection in big data sets. Density based DBSCAN (Density-based spatial clustering of applications with noise) algorithm from clustering algorithms which are sensitive to noisy and outlier values is one of the most important methods in outlier detection. In this study, an application was developed using DBSCAN algorithm in C# programming language for the detection of outliers. In the developed application, 2 data sets with different data numbers were examined and analyzed. For the shortest possible data analysis time, serial and parallel programming techniques were used separately. In order to shorten the analysis time of big data sets, parallel class members in TPL (Task Parallel Library) provided with .Net 4.0 were used. In series of analysis of data sets, it was seen that DBSCAN algorithm produces more accurate results and is more practicable than other selected algorithms in terms of outlier detection. When considered in terms of computing performance, parallel programming has become more efficient as the number of data increases. [ABSTRACT FROM AUTHOR]

It is so important to determine outlier, influence and leverage points in multiple linear regression analysis for the accuracy of statistical inferences. To detect the influence observations, Nurunnabi et al. (2016) proposed a robust influence distance (ID). However, the determination of observations that would not be used in the calculations of this distance are based on non-robust statistics. Thus, it is affected by outliers. In this paper, it is suggested that influence distance based on robust estimators (RID) could be used for detecting influence observations. Moreover ID and RID's which were used to determine outliers, are applied to two known data sets and are compared based on simulation studies. The results show that RID based on RLS performs the best. [ABSTRACT FROM AUTHOR]

Factor analysis, one of those multivariate methods, is used for data reduction, determining the relationships between variables, and also as a classification method. The researchers frequently face with the observations that might have bad affects on the results while analysing data sets with many variables. Those kind of observations are called outliers. Outliers, usually evaluated as erroneous observations, may reflect important information about the data set, but on the other hand may cause misspecification of the model, biased parameter estimates, and may lead incorrect analysis results. The aim of this study is to use factor analysis method which will not make biased parameter estimations of the outliers. For this aim, among the robust local and scale parameter estimators which reduce the affects of outliers, fast minimum covariance determinant estimators are choosen and an application is performed. As a result, robust factor analysis estimation results are obtained which are mostly fitted to the data by reducing the affect of the outliers, with higher explained variability and with explained variables which are gathered on more significant factors. [ABSTRACT FROM AUTHOR]

Regression analysis is one of the most commonly used statistical techniques. Out of many possible regression techniques, the Least Squares Method (LSM) has been generally adopted because of tradition and ease of computation. In data analysis and trend modelling applications the least squares (LS) estimator is widely used and LS regression is, in most cases, the method of choice. However, the crucial fact that the LS estimator is very sensitive to outlying observations may lead to unreliable results in the regression estimates and, hence, to a misleading interpretation of the data. To remedy this problem, some statistical techniques have been developed that are not so easily affected by outliers. These are the robust methods, the results of which remain trustworthy even if a certain amount of data is outlier. One of them is the least median squares method which is using in statistical analysis. In this study, estimation of Least Square and Least Median Square has been given. LS and LMS methods are applied and compared on differrent sample that can be produced by simulation study. To find whether there is important difference between methods are compared their estimations ((β̂0, β̂1, σ̂2, R2) [ABSTRACT FROM AUTHOR]