kNN Classification with an Outlier Informative Distance Measure

Bhattacharya, Gautam; Ghosh, Koushik; Chowdhury, Ananda S.

doi:10.1007/978-3-319-69900-4_3

Cited by 8 publications

(4 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Finally, the sensitivity analysis further proves that the K-NN classifier can be exploited only when a limited amount of data is incorrectly labeled or in the case that a limited number of outliers is included within the dataset [44]. In fact, both accuracy and posterior probability of the K-NN classifier significantly decrease when the rate of UMI is approximately equal to 20%, so that only 66% of data are correctly labeled.…”

Section: Sensitivity Analysismentioning

confidence: 75%

Optimal Classifier to Detect Unit of Measure Inconsistency in Gas Turbine Sensors

et al. 2022

View full text Add to dashboard Cite

Label noise is a harmful issue that arises when data are erroneously labeled. Several label noise issues can occur but, among them, unit of measure inconsistencies (UMIs) are inexplicably neglected in the literature. Despite its relevance, a general and automated approach for UMI detection suitable to gas turbines (GTs) has not been developed yet; as a result, GT diagnosis, prognosis, and control may be challenged since collected data may not reflect the actual operation. To fill this gap, this paper investigates the capability of three supervised machine learning classifiers, i.e., Support Vector Machine, Naïve Bayes, and K-Nearest Neighbors, that are tested by means of challenging analyses to infer general guidelines for UMI detection. Classification accuracy and posterior probability of each classifier is evaluated by means of an experimental dataset derived from a large fleet of Siemens gas turbines in operation. Results reveal that Naïve Bayes is the optimal classifier for UMI detection, since 88.5% of data are correctly labeled with 84% of posterior probability when experimental UMIs affect the dataset. In addition, Naïve Bayes proved to be the most robust classifier also if the rate of UMIs increases.

show abstract

Section: Sensitivity Analysismentioning

confidence: 75%

Optimal Classifier to Detect Unit of Measure Inconsistency in Gas Turbine Sensors

et al. 2022

View full text Add to dashboard Cite

show abstract

“…Bhattacharya dkk. [6] menyebutkan keberadaan outlier pada dataset dapat mempengaruhi akurasi algoritme kNN. Lebih lanjut, kajian tersebut menyatakan bahwa skor outlier berdasarkan perbedaan tingkat kepadatan dapat digunakan untuk memodulasi fungsi jarak potensial pada klasifikasi kNN dalam upaya meningkatkan hasil akurasi pada dataset yang memiliki outlier.…”

Section: Pendahuluanunclassified

“…Berbeda dengan [6] dan [7] dalam meningkatkan hasil akurasi klasifikasi, penelitian ini mengusulkan prapemrosesan untuk menghilangkan outlier pada dataset untuk meningkatkan hasil akurasi klasifikasi algoritme kNN. Berbeda dengan [10] dan [11] dalam menerapkan algoritme K-means dalam mendeteksi outlier, penelitian ini mengusulkan deteksi outlier dengan menggunakan algoritme K-means dan matriks jarak pada dataset yang sudah memiliki kelas label.…”

Section: Pendahuluanunclassified

Preprocessing kNN algorithm classification using K-means and distance matrix with students’ academic performance dataset

Sugriyono

Siregar

2020

Jurnal Teknologi dan Sistem Komputer

View full text Add to dashboard Cite

The existence of outliers in the dataset can cause low accuracy in a classification process. Outliers in the dataset can be removed from a preprocessing stage of classification algorithms. Clustering can be used as an outlier detection method. This study applies K-means and a distance matrix to detect outliers and remove them from datasets with class labels. This research used a dataset of students’ academic performance totaling 6847 instances, having 18 attributes and 3 class labels. Preprocessing applies the K-means method to get centroid in each class. The distance matrix is used to evaluate the distance of instance to the centroid. Outliers, which are a different class, will be removed from the dataset. This preprocessing improves the classification accuracy of the kNN algorithm. Data without preprocessing has 72.28 % accuracy, preprocessed data using K-means with Euclidean has 98.42 % accuracy (an increase of 26.14 %), while the K-means with Manhattan has 97.76 % accuracy (an increase of 25.48 %).

show abstract

“…The advantages of this algorithm include fast training, simple and easy to learn, and effective in training big data [8]. However, KNN can be affected by outliers, as a result, its accuracy is not good enough [9]. Another method used in KNN is Subspace Outlier Detection (SOD), which is an outlier detection method that models outliers on high-dimensional data.…”

Section: Introductionmentioning

confidence: 99%

The Implementation of Subspace Outlier Detection in K-Nearest Neighbors to Improve Accuracy in Bank Marketing Data

Anggoro¹

2020

IJETER

View full text Add to dashboard Cite

Banks are the most important institution in the economics and social sector. Hence, it provides services or products such as term deposits. Term deposits give banks the most reliable source of profit and credits. Bank uses direct marketing to attract customers to subscribe to the deposit program. By using data mining, a bank can predict the client who will subscribe to the term deposit program. In this research, the dataset to be tested is bank marketing by Portuguese banking institutions. The data mining algorithm used is K-Nearest Neighbors (KNN), but it has a sensitive weakness to outliers which causes the accuracy of this algorithm is not good. This research aims to cover up the weakness of the KNN by detecting outliers in the dataset and improving the accuracy of the KNN. The outlier detection method used in this research is Subspace Outlier Detection (SOD) and Principal Component Analysis (PCA) as dimensional reduction method. The dataset was splitting into 70% training data and 30% testing data as new data. K-fold cross-validation was used to search the K Neighbors value of KNN during the modeling training data. The classification results of testing data based on the confusion matrix show the accuracy of the KNN algorithm which only uses PCA is 91%. While the KNN algorithm accuracy with SOD and PCA is 94%. It can be concluded that the accuracy of KNN using SOD is better than accuracy of KNN without SOD by improving the accuracy by 3% in the bank marketing dataset.

show abstract

kNN Classification with an Outlier Informative Distance Measure

Cited by 8 publications

References 12 publications

Optimal Classifier to Detect Unit of Measure Inconsistency in Gas Turbine Sensors

Optimal Classifier to Detect Unit of Measure Inconsistency in Gas Turbine Sensors

Preprocessing kNN algorithm classification using K-means and distance matrix with students’ academic performance dataset

The Implementation of Subspace Outlier Detection in K-Nearest Neighbors to Improve Accuracy in Bank Marketing Data

Contact Info

Product

Resources

About