Missing Values Imputation using Feed Forward Neural Network
Saravanan P1, Justin Samuel S2, Nirmalrani V3, Mathivanan G4
1Saravanan P, Research Scholar, School of Computing, Sathyabama Institute of Science and Technology, Chennai, India.
2Justin Samuel,Department of Computer Science and Engineering, PSN College of Engineering and Technology, Tirunelveli, India.
3Nirmalrani V, School of Computing, Sathyabama Institute of Science and Technology, Chennai, India.
4Mathivanana G, School of Computing, Sathyabama Institute of Science and Technology, Chennai, India.
Manuscript received on 6 August 2019. | Revised Manuscript received on 12 August 2019. | Manuscript published on 30 September 2019. | PP: 641-643 | Volume-8 Issue-3 September 2019 | Retrieval Number: B2503078219/19©BEIESP | DOI: 10.35940/ijrte.B2503.098319
Open Access | Ethics and Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: Data cleaning intends to make the data as highly qualified one in terms of completeness and noisy free to makes the pattern outcome as high quality. Since most of the real world data possess these crucial issues, finding the most probable value for the holes introduced in data collection becomes a challenging task. This paper attempts to employ the feed forward neural network to make the collected dataset as complete which in turn the pattern outcome also complete. The collected dataset which possess the missing values is used to generate the identity matrix where the filled cells might get one and rest of the cells as zero. The given dataset gets normalized using minmax variety after replacing the missing cells as zero which will become a target matrix. By adjusting the weight values for the edges across the various edges the net value gets computed. The process gets repeated with a small increment done over the input to reach the target till the loss function yields the desirable value. The method is experimented with various UCI machinery dataset for different standard missing ratios. The proposed system performance is evaluated through RMSE parameter and the above method shows better accuracy with other popular methods.
Keywords: Data Cleaning, Missing Values, Imputation, RMSE, Pattern Discovery.
Scope of the Article: Neural Network