Informatics Publishing Limited

V. Deneshkumar ¹, K. Senthamarai Kannan ², M. Manikandan ²

Affiliations
1 Department of Statistics, Manonmaniam Sundaranar University, Tirunelveli-627012, IN
2 Department of Statistics, Manonmaniam Sundaranar University, Tirunelveli, IN

Abstract

Outlier detection is important in many fields. In statistics, an outlier is a observation that is numerically far-away from the rest of the data. The handling of outlying observations in a data set is one of the most important tasks in data pre-processing. The large data base can be classified in an unsupervised manner using clustering and classification algorithms. Fuzzy C-means is a method of clustering which was developed by Dunn in (1973) and improved by Bezdek in (1981). This allocates one piece of data in two or more clusters and it is frequently used in pattern recognition. Herein a proposed method based on Fuzzy approach which combines outlier analysis and clustering technique is presented. Clustering validation technique adaptively evaluated the results of a clustering algorithm. A numerical example is provided for illustration using iris data set.

Keywords

Outlier Detection, Fuzzy Clustering, Silhouette Index, FCM Algorithm, and Random Number Simulation.

Full Text

     

V. Deneshkumar ¹, K. Senthamarai Kannan ²

Affiliations
1 Department of Statistics, Manonmaniam Sundaranar University, Tirunelveli-627012, IN
2 Department of Statistics, Manonmaniam Sundaranar University, Tirunelveli, IN

Abstract

For many data mining applications, finding the outliers is more interesting than finding the common patterns of the data. Outliers are frequently adapted in time series data mining analysis. The main objective of this paper, outliers on forecasting in agricultural production is analyzed. Outliers in time series data was carried out by Fox (1972). Outlier detection has been used for detect and, where appropriate, remove inconsistent observations from data. The original outlier detection methods were arbitrary but new, Principled and systematic techniques are used, drawn from the full scope of computer science and statistics. In agricultural production outliers are initially detected and then forecast using ARIMA model. Predictions made after detecting outliers are compared with numerically and graphically the predictions made before detecting outliers.

Keywords

Data Mining, Outliers, Forecasting, Mean Square Error and ARIMA Model.

Full Text