Data Mining Classification Models for Industrial Planning

The data mining models are an excellent tool to help companies that live from the sale of items they produce. With these models combined with Lean Production, it becomes easier to remove waste and optimize industrial production. This project is based on t

  • PDF / 128,922 Bytes
  • 10 Pages / 439.37 x 666.142 pts Page_size
  • 77 Downloads / 224 Views

DOWNLOAD

REPORT


goritmi Research Centre, University of Minho, Guimarães, Portugal [email protected], {cfp,mfs}@dsi.uminho.pt 2 Value Added Partners, Porto, Portugal

Abstract. The data mining models are an excellent tool to help companies that live from the sale of items they produce. With these models combined with Lean Production, it becomes easier to remove waste and optimize industrial production. This project is based on the phases of the methodology CRISP-DM. Several methods were applied to this data namely, average, mean and standard deviation, quartiles and Sturges rule. Classification Techniques were used in order to understand which model has the best probability of hitting the correct result. After performing the tests, model M1 was the one with the best chance to accomplish a great level of classification having 99.52% of accuracy. Keywords: Data mining  Classification  CRISP-DM  DSR  Lean  WEKA

1 Introduction Companies in industry are increasingly feeling the need to find a way to optimize their production to meet the adversities of the economic world. One of the best ways to do this is using the data mining models that allow, based on past sales, get an estimate of how much will sell the right time, production efficiently and reducing the waste of raw material and labor work. The rating models are a great tool to help businesses achieved success. Having said that, test yourself several models in terms of classification hoping to find models with higher accuracy than 90%. This paper initially made an overview of the concepts inherent in the project including data mining, Lean Production and Decision Support ending with what exists today. Subsequently it is presented methodologies, CRISP-DM, DSR and at the level of the tools was used WEKA. Then the paper presents the results following the phases of the CRISP-DM. Finally, it created a little discussion about the project and concluding with the presentation of closing arguments.

© Springer Nature Singapore Pte Ltd. 2017 M. Singh et al. (Eds.): ICACDS 2016, CCIS 721, pp. 585–594, 2017. DOI: 10.1007/978-981-10-5427-3_60

586

R. Bragança et al.

2 Background 2.1

Data Mining

In the world there is a vast amount of data stored, but these are only examined in a very superficial way, which leads to having a wealth of data and great poverty of knowledge [1]. In the last decades, data mining has been widely recognized as an powerful and versatile tool of data analysis and has made a significant contribution in the areas of information technology clinical medicine, sociology, physics, in the areas of management, economics and finance [2]. Data mining (DM) is the task that seeks to discover patterns in data sets by using methods of artificial intelligence, machine learning, statistics and database systems [3]. Wu [4] suggests a more complete definition for DM, he believes it is the integration of various subjects: (i) databases, (ii) databases technologies, (iii) statistic, (iv) machine learning, (v) math, (vi) neural networks and others. The DM was designed to use two techniques