A hybrid prediction model of Artificial Neural Network (ANN) and Rough Set Theory (RST) is proposed in this paper. ANN and RST are frequently applied to different data mining problems. Hybrid approach of combining ANN and RST can be directly applied to classification and regression without additional transformation mechanisms in the data set. ANN is one of the most powerful and universal predictors. On the other hand, the use of RST in knowledge extraction field is proved by the growing number of applications and publications. The main advantage of RST is the non-requirement of prior information about the data for the selection of reduct set. In this paper, a new Rough Neural Network (RNN) algorithm for the proposed hybrid approach in order to predict semen fertility rate is described. This algorithm is composed of two parts namely pre-processing part based on RST and classification part based on ANN. However, there arises an issue of selecting the most significant parameters from the In Vitro Fertility (IVF) tests outcomes.

Prediction of sperm fertility rate has great economic importance for breeding animals when Artificial Insemination (AI) is used, and AI is one of the most successful reproductive technologies developed to improve reproductive efficiency of farm animals and dairy cattle [1]. The basic purpose of semen evaluation procedures is to ensure that only good quality and highly fertile semen is used for AI purposes. Therefore, it is of utmost important to ensure the accurate evaluation of a bulls fertility since a single ejaculate provides several insemination doses and shows much influence on the reproductive potential of a herd [2]. Since fertilization is a complex process, semen samples can be subjected to various in vitro tests that are related to the process of fertilization. Traditionally, statistical techniques of linear and nonlinear regression are used to predict fertility rate of sperm. The accurate prediction of semen fertilization potential is very essential and suitable mathematical techniques or models employed to predict fertility rate play a crucial role. Machine learning techniques, such as ANN and RST, might further improve predictive models of semen fertilization potential. ANNs are mathematical models which have the capability of relating the input and output parameters, and learning from examples through iteration, without requiring a prior knowledge of the relationships between the process parameters. This paper presents a hybrid approach of combining rough set and neural network methods for decision and classification support. Both methodologies have their place among intelligent classification and decision support methods [3].

RST is a mathematical approach proposed by Z Pawlak and has come into focus as an alternative to the more widely used method of machine learning and statistical data analysis [4-5]. The data set represented in the RST framework is in the form of information system or table. Each row of the information system represents an object and every column represents an attribute that can be measured for each object. Rough set theory is derived from set theory. Therefore, usual assumptions of traditional quantitative research techniques do not apply.

Science and Technology Journal, Neural Networks, Rough Set Theory, Artificial Neural Network, Rough Neural Network Algorithm, Artificial Insemination, Nonlinear Regressions, Neural Network Methods, Decision Support Methods, Information Systems, Quantitative Research Techniques, Standard Programming Techniques, Spermatological Data.