Hybrid Kernel Fuzzy Clustering with Feed Lion Neural Network for Missing Data Imputation and Classification
Journal Title: International Journal of Computer Science Engineering and Information Technology Research (IJCSEITR) - Year 2017, Vol 7, Issue 5
Abstract
A common issue in many practical applications associated with pattern classification is data incomplete or missing data due to various reasons that differ based on the applications. Missing data imputation is a promising approach used to handle this issue, which fills the missing attributes with estimated values following several techniques. This paper proposes a new strategy for data imputation and classification using a hybrid prediction model that combines Hybrid Kernel Fuzzy Clustering (HKFC) and Feed Lion Neural Network (FLNN). FLNN considers the missing attributes as the class attributes to predict the missing data, which will also be computed using HKFC and finally, the combined effect of these two approaches provides the missing data. FLNN is designed for the classification by modifying Leven berg-Marquardt (LM) - based feed forward neural network with the incorporation of LOA and thus, selects the weights optimally. The experiment is carried out using three data sets from UCI, based on two metrics, Mean Squared Error (MSE) and accuracy. The performance of the proposed method, HKFC+FLNN is compared with three techniques, FC+FLNN, FC+KNN and K-Means+KNN, where HKFC+FLNN attained MSE by 0.0344, 0.0048, 0.2754, and classification accuracy of 0.8022, 0.95 and 0.9741, for heart disease, iris, and wine datasets, respectively.
Authors and Affiliations
R Rajani, T. Sudha
Keywords
Related Articles
- EP ID EP244999
- DOI -
- Views 116
- Downloads 0
E2TBR: Energy Efficient Transmission Based Routing for IoT Networks
The Internet of Things (IoT) envisions the notion of ubiquitous connectivity of everything. This combines many technologies. Since millions of devices are going to be connected to the internet and wireless in nature, ene...
Automated Hydroponics System using Nft System and IOT
In the urban areas there is lack of open green space for farming and even if the land is available it is infertile for plants to grow on them. Problems faced in urban areas farms are due to the toxic elements let in the...
Local Binary Patterns Based Detection of Rust Disease of Lentils (Lens culinaris) Using k-NN Classification System
This research paper reported the role of k-Nearest Neighbor (k-NN) classifier for detection and classification of rust disease of Lens culinaris at microscopic level, which is a very initial stage of disease i.e. haustor...
The Scheme Programming Language
This paper is a review of the Scheme programming language. Origin of Scheme, various features, syntax, example code and applications of Scheme language are discussed in this paper. The interpreters that are used and othe...
Efficient Traffic Management with Intermediate Data Partition for Big Data Applications33
The MapReduce programming models improves huge scale information handling on ware bunch by exploiting parallel map tasks and reduce tasks. Although many numerous endeavors have been made to enhance the execution of Map R...