International Journal of Engineering and Advanced Technology (IJEAT) ISSN: 2249 – 8958, Volume-9 Issue-1, October 2019 1778 Published By: Blue Eyes Intelligence Engineering & Sciences Publication Retrieval Number: F9060088619/2019©BEIESP DOI: 10.35940/ijeat.F9060.109119 Early Detection of Diabetic Mellitus Based On Modeling Techniques Pydipala Laxmikanth, Bhramaramba Ravi Abstract: Rapid growth of population in particular in elderly, signifies the issues of healthcare have become a concern. Also the lifestyle changes together with social and economical factors influences the cause of disease generation among the generated diseases, diabetic disease is mostly populated. Therefore effective measures are to be taken so that the early wakeup with regard to disease treatment helps to minimize the after effects. In this article, a novel methodology based on Gaussian mixture model built for analyzing the patients and help to identify the disease during the primitive stage. The methodology is presented based on PIMA INDIAN DATASET.The Results derived showcase efficacy of the developed method. Index Terms: Diabetes mellitus, Gaussian mixture model, disease detection, probabilistic method, early identification I. INTRODUCTION Diabetes Mellitus is a kind of metabolic disease wherein the individuals suffering with this disease posses high Blood glucose. This may be due to lack of insulin production or may be due to poor response of the human cells towards insulin and thus causes the individuals with symptoms like frequent urination, thirstiness and hungriness. Most of the literature in this area have classified this disease as a metabolic disorder [1][2][3]. The conditions that trigger are mostly due to radical increase levels of insulin or decrease in insulin levels. Generally, once the disease identified, it is uncontrollable, and leads towards further complications which may lead towards mortality [4][5]. Frequently, used terminologies with respect to this disease classify it into two types, type1 and type2. In typical cases, wherein the conditions of the human body does not generate adequate hormones which allow to absorb and utilize the glucose levels, it treated as type1 disease and hormone deficiency is considered to be insulin [6] insulin resistance is generally considered as a pre-cursor for type2 diabetes and in most of the cases the factors that are assumed to have an impact includes overweight and insufficient or inactive lifestyle. Type1 diabetes leaves no clues for prompting and in most of the cases. To prevent the disease, insulin is to be advocated at regular intervals. The main impacts of improper treatment and care leads towards critical diseases like; retinopathy, neuropathy, nephropathy and even cardiac diseases [7][8][9]. In contrary type2 diabetes is considered to be less dangerous, if proper care is taken. Revised Manuscript Received on October 20, 2019. The cases with respect to these disease are mostly increasing not only in India but also it is observed that an overall of 20 % growth rate is observed worldwide every year in the number of patients suffering with this disease. Therefore effective steps are to be taken to identify the disease at the earlier stage and plan effective measures such that, the after impact can be minimized. Lot of activate research has been witnessed in this area and most of the works presented by various authors focus on either identifying the sensitivity of insulin and understanding the impact of exercises on plasma glucose and insulin levels [10][11]. Models are also presented for critically examining in the patients of type1 and type2 individually and suggestive methods are there by offered with regard to the physical exercises [12]. methodologies based on data mining techniques like; decision trees, Bayesian, k-nearest neighbor also highlighted in the literature to understand the impact of the disease and measures to be taken to narrow down its impact onto the other organs [13][14][15]. Models based on neural networks, optimization techniques are also quoted [16][17][18][19]. However, in majority of the works presented, least importance is given towards the early detection of the disease rather than combating the after effects of the disease. Therefore in this article, methodology is proposed which aims at early detection of the disease using Gaussian mixture models [20]. The main objective behind the choice of Gaussian mixture models includes its abilities to identify the abnormal increase of blood sugar levels having complex densities. By adjusting the means and covariance’s, linear combinations can be generated with varying levels of blood levels, which helps to identify the early impact of the disease during the start stage. The remaining part of the paper is articulated in the following fashion: In section 2, the Gaussian mixture model is presented. The section 3 highlights the article by generating an alert condition. In section 4, the dataset considered is presented. The section 5 of the article, deals with the experimentation. In section 6, the results derived are analyzed. The concluding section 7 summarizes the article. II. GAUSSIAN MIXTURE MODELS It is a probabilistic methodology which assumes that the data is modeled using a finite mixture of Gaussian distributions. It generally considers the Bayesian densities. In one way it is assumed to be a generalization of k-means clustering coupled the covariance structure and about the cluster centers. It has a basic significant property where in case of complex data, sufficient number of Gaussian can be generated and then means and covariance’s together with them linear combination can be approximated most proximately. The most specific symptoms are considered and