Determining the representative features of polycystic ovary syndrome via Design of Experiments

Polycystic ovary syndrome(PCOS) is the most common syndrome found in women around the world. In India, earlier one in every ten women had PCOS and nowadays, one in every five women has PCOS. It is not a disease but it’s a syndrome as it may lead to several diseases like diabetes, high blood pressure, irregular periods, etc. PCOS symptoms vary according to age like in teenagers, PCOS may detect by irregular periods and in middle age, PCOS may detect by infertility, cancer in the uterus, risk of miscarriages, etc. In the modern era, PCOS identification is dependent on various parameters. In this article, the objective is to find the most important parameters for identifying the PCOS with the Design of experiments (DOE) analysis tool, we want to answer the question “Can we identify the PCOS with less number of parameters”. The 2 k-p fractional factorial design was utilized as it can take a large number of inputs and gives the response in a fewer number of experiments. PCOS dataset of 541 instances and 7 attributes were taken to implement DOE in which 7 attributes are reduced to 4 attributes. The percentage of variation of reduced attributes, Follicle No. (r), FSH/ LH, Follicle No. (l), Skin Darkening has been obtained as 28.44%, 21.36%, 15.29% and 15.29% respectively through 2 k-p fractional factorial design method. DOE response was validated using ANOVA and Minitab statistical software. Pareto chart reference line indicates the effectiveness of Follicle No. (r), FSH/ LH, Follicle No. (l) and Skin Darkening. This analysis can help doctors to diagnose PCOS and researchers to save time by disposing of irrelevant parameters when performing experiments for PCOS diagnosis-related studies in future.