Storage time prediction of pork by Computational Intelligence Ana Paula A.C. Barbon a,⇑ , Sylvio Barbon Jr. b,1 , Rafael Gomes Mantovani c , Estefânia Mayumi Fuzyi b , Louise Manha Peres a , Ana Maria Bridi a a Department of Zootechnology, Londrina State University (UEL), Londrina 86057-970, Brazil b Department of Computer Science, Londrina State University (UEL), Londrina 86057-970, Brazil c Sciences Institute of Mathematics and Computers (ICMC), University of São Paulo (USP), São Carlos, Brazil article info Article history: Received 3 February 2016 Received in revised form 6 June 2016 Accepted 23 June 2016 Available online 4 July 2016 Keywords: Fuzzy rule based system Machine learning Meat quality Classification Post mortem abstract In this paper, a storage time prediction of pork using Computational Intelligence (CI) model was reported. We investigated a solution based on traditional pork assessment towards a low time-cost parameters acquisition and high accurate CI models by selection of appropriate parameters. The models investigated were built by J48, Naïve Bayes (NB), k-NN, Random Forest (RF), SVM, MLP and Fuzzy approaches. CI input were traditional quality parameters, including pH, water holding capacity (WHC), color and lipid oxida- tion extracted from 250 samples of 0, 7 and 14 days of post mortem. Five parameters (pH, WHC, L / , a / and b / ) were found superior results to determine the storage time and corroborate with identification in min- utes. Results showed RF (94.41%), 3-NN (93.57%), Fuzzy Chi (93.23%), Fuzzy W (92.35%), MLP (88.35%), J48 (83.64%), SVM (82.03%) and NB (78.26%) were modeled by the five parameters. One important obser- vation is about the ease of 0-day identification, followed by 14-day and 7-day independently of CI approach. Result of this paper offers the potential of CI for implementation in real scenarios, inclusive for fraud detection and pork quality assessment based on a non-destructive, fast, accurate analysis of the storage time. Ó 2016 Elsevier B.V. All rights reserved. 1. Introduction The parameters involving meat quality are of most importance for the meat processing industry. Research projects are often developed to assess improvements in measurements and quality assessments, as well as factors that influence pork quality such as environmental conditions, pre-slaughter management and pur- chasing decisions of consumers (Rosenvold and Andersen, 2003). A perspective of pork quality evaluation is based on the muscle to meat conversion. This process includes several enzymatic and pro- tein denaturation processes that directly influence pH and other quality attributes (Salmi et al., 2012). These parameters are play a major role on quality and are related to post mortem period, mainly because the rate of glycolysis, affecting the technological quality of meat (Hammelman et al., 2003). Determination of post mortem period is relevant for the indus- try because it allows identification of aging period and freshness evaluation, as well as indicating the consumer preferences. Another advantage is to identify potential fraud during the food storage period. Besides, it can indicate storage problems as temper- ature deviation in cold rooms, freezers, and refrigerators that can lead to meat deterioration and shelf-life reduction. Nowadays, consumers are more demanding for food quality, as they are looking for clear and reliable information about product origin, production method and food preservation (Sentandreu and Sentandreu, 2014). Fraud in the meat sector is constantly described and (Ballin, 2010) describes that fraud can be categorized according to the pos- sibility of occurrence: origin of meat, meat replacement, and meat processing. Moreover, within each of these frauds there are subcat- egories: post mortem period, meat cuts, animal breed, meat fresh- ness, among others. However, during post mortem period, some meat quality parameters may be modified, e.g. pH, Water Holding Capacity (WHC), color and lipid oxidation (Tarsitano et al., 2013). The Meat freshness determines the choice of the product by the consumer (Xiong et al., 2015). Moreover, this assessment is also measured by quality parameters mentioned before, and depends directly on the storage time. Nonetheless, laboratory evaluation parameters are costly, time consuming, dependent on trained persons and sub- jective evaluation. In this context arise alternative methods and non-destructive analysis of food using computational tools (Chen et al., 2011). http://dx.doi.org/10.1016/j.compag.2016.06.028 0168-1699/Ó 2016 Elsevier B.V. All rights reserved. 1 http://www.uel.br/grupo-pesquisa/remid/. ⇑ Corresponding author. E-mail addresses: apbarbon@gmail.com (A.P.A.C. Barbon), barbon@uel.br (S. Barbon Jr.), rgmantov@icmc.usp.br (R.G. Mantovani), emfuzyi@uel.br (E.M. Fuzyi), louise_mp@zootecnista.com.br (L.M. Peres), ambridi@uel.br (A.M. Bridi). Computers and Electronics in Agriculture 127 (2016) 368–375 Contents lists available at ScienceDirect Computers and Electronics in Agriculture journal homepage: www.elsevier.com/locate/compag