Search

Pertumbuhan data yang terjadi saat ini berpengaruh terhadap analisis data di berbagai bidang, seperti astronomi, bisnis, kedokteran, pendidikan, dan finansial. Data yang terkumpul dan tersimpan mengandung nilai ekstrem atau nilai pengamatan yang berbeda dari kebanyakan nilai hasil pengamatan lain. Nilai ekstrem tersebut disebut dengan outlier. Outlier pada sebagian data sering kali memiliki nilai yang mengandung informasi penting, sehingga perlu dikaji agar dapat diambil keputusan untuk menghapus atau menggunakan data tersebut sebelum diterapkan dalam penambangan data. Deteksi outlier dapat dilakukan sebagai prapemrosesan data dengan menggunakan teknik analisis outlier. Beberapa teknik analisis outlier yang banyak diterapkan antara lain metode berbasis jarak (distance), metode berbasis kepadatan (density), dan metode local outlier factor (LOF). K-nearest neighbors (KNN) merupakan salah satu algoritma penambangan data yang sangat sensitif terhadap outlier karena cara kerjanya yang bergantung pada nilai k. Oleh karena itu, perlu penanganan tepat saat KNN bekerja pada dataset yang mengandung outlier. Metode eksperimen dipilih dalam menerapkan metode usulan, dengan tujuan untuk mengoptimasi algoritma KNN berdasarkan perbandingan analisis outlier (KNN-distance, KNN-density, dan KNN-LOF). Hasil penelitian menunjukkan bahwa KNN-kepadatan unggul sebanyak tiga kali: pada Wisconsin Breast Cancer dengan nilai rata-rata akurasi sebesar 99,34% pada k=3 dan k=5; pada Glass dengan nilai rata-rata akurasi sebesar 85,25% pada k=7; dan pada Lymphography dengan nilai rata-rata akurasi sebesar 85,45% pada k=5. Selanjutnya, berdasarkan hasil uji Friedman dan uji Nemenyi, juga terbukti bahwa ada perbedaan yang signifikan antara KNN-kepadatan dengan KNN-LOF.

Poverty is defined as an inability of the individual to meet basic needs for a decent life. According to BPS data in 2020, Sulawesi Island ranks fifth as the poorest island in Indonesia. This study aims to find out the mapping of areas and indicators of poverty in Sulawesi Island using Biplot Analysis with Robust Singular Value Decomposition approach for outlier research data. Based on the results of the study, there are five objects that are outlier and the information provided by the biplot amounted 98.45%. District/city that have similar characteristics are divided into 4 groups. The indicator of poverty that has the most diversity is the School Old Expectations Numbers (Var 4) and the one with the least diversity is Poor Households Using Clean Water (Var 8). Indicators of poverty that are positively correlated are Literacy Numbers (Var 1) and Non-Working Poor Population (Var 5), while the negative correlated are The Non-Working Poor Population (Var 5) and Poor Households Using Clean Water (Var 8). There are 19 districts/cities that have literacy values above the average of all districts/cities and 11 districts/cities that have a per capita expenditure value below the average of all districts/cities.

Statistical analysis which aims to analyze a linear relationship between the independent variable and the dependent variable is known as regression analysis. To estimate parameters in a regression analysis method commonly used is the Ordinary Least Square (OLS). But the assumption is often violated in the OLS, the assumption of normality due to one outlier. As a result of the presence of outliers is parameter estimators produced by the OLS will be biased. Bootstrap Residual is a bootstrap method that is applied to the residual resampling process. The results showed that the residual bootstrap method is only able to overcome the bias on the number of outliers 5% with 99% confidence intervals. The resulting parameters estimators approach the residual bootstrap values ??OLS initial allegations were also able to show that the bootstrap is an accurate prediction tool.

The presence of outliers in observation can result in biased in parameter estimation using ordinary least square (OLS). Robust regression MM-estimator is one of the estimations methods that able to obtain a robust estimator against outliers. Random sample consensus (ransac) is another method that can be used to construct a model for observations data and also estimating a robust estimator against outliers. Based on the study, ransac obtained model with less biased estimator than robust regression MM-estimator.

Multiple linear regression analysis with a lot of independent variable always makes many problems because there is a relationship between two or more independent variables. The independent variables which correlated each other are called multicollinearity. Principal component analysis which based on variance covariance matrix is very sensitive toward the existence of outlier in the observing data. Therefore in order to overcome the problem of outlier it is needed a method of robust estimator toward outlier. ROBPCA is a robust method for PCA toward the existence of outlier in the data. In order to obtain the robust principal component is needed a combination of Projection Pursuit (PP) with Minimum Covariant Determinant (MCD). The results showed that the ROBPCA method has a bias parameter and Mean Square Error (MSE) parameter lower than Principal Component Regression method. This case shows that the ROBPCA method better cope with the multicollinearity observational data influenced by outlier.

Principal Component Regression is a method to overcome multicollinearity techniques by combining principal component analysis with regression analysis. The calculation of classical principal component analysis is based on the regular covariance matrix. The covariance matrix is optimal if the data originated from a multivariate normal distribution, but is very sensitive to the presence of outliers. Alternatives are used to overcome this problem the method of Least Median Square-Minimum Covariance Determinant (LMS-MCD). The purpose of this research is to conduct a comparison between Principal Component Regression (RKU) and Method of Least Median Square - Minimum Covariance Determinant (LMS-MCD) in dealing with outliers. In this study, Method of Least Median Square - Minimum Covariance Determinant (LMS-MCD) has a bias and mean square error (MSE) is smaller than the parameter RKU. Based on the difference of parameter estimators, still have a test that has a difference of parameter estimators method LMS-MCD greater than RKU method.

There are several problems in industrial process for example problems associated with product quality. In statistics, observation which is significantly different to the average is called outlier. The outlier can give significant influence to the result of modeling, which can affect the decision making. This research develops the outlier detection method using the Likelihood Displacement Statistic method, called Likelihood Displacement Statistic-Lagrange (LDL) method. The LDL method is applied to sugar and molasses production data of Djombang Baru Sugar Factory, Jombang, East Java. The result of this research shows that factors influenced the sugar and molasses production are sugar cane with the dirt less than 5%, sugar cane with the dirt between 5% to 7%, sugar cane with the dirt higher than 7%, and imbibition water