This project is related for prediction the diabetes in patients, based on some medical features the bulit model used to classify the person will effected with diabetes or nor. The model was bulid with various techniques which are base model evaluation, random search CV and nature inspired algorithm. The detailed mathematical equations related to nature inspired algorithms were discussed in PPT. All the files which are requried to carry out this project were added in repository which includes data sets, codeing, and user inteface files. The UI has been build with the help flask server.
A significant variety of nature-inspired techniques are used in the field of Machine Learning (ML) to handle a wide range of problems. The term ”evolutionary computation” refers to a group of global optimization algorithms inspired by biological evolution.They’re a type of population-based trial-and-error problem solver that includes a metaheuristic or stochastic optimization function. The aim of this work is to create a model that can effectively detect diabetes from the given dataset using nature inspired algorithms. This work implants the usage of nature inspired optimization technique for predicting diabetes. The nature inspired algorithms such as bat algorithm, hybrid bat algorithm, grey wolf optimization algorithm and firefly algorithm were mostly used for numerical data and also used for improving the accuracy and other performance metrics. Seven classification algorithms were used in detection of diabetes and the Nature inspired algorithms are used to tune the hyperparameters of the classifiers in order to get an enhanced accuracy. Majority of the used classifiers’ performance was enhanced with the use of nature inspired algorithms. Firefly algorithm and hybrid bat algorithm outperforms the other algorithms with the highest accuracy of 77.2% in the case of random forest classifier.
Here, I have used other classification algorithms like gaussianNB, svm, extra trees classifier and K-Nearest Neighbour along with the existing methods. Apart from these we further used two more technique’s for predicting the model by hyper-parameter tuning using randomized search and nature inspired algorithm. Based on these three approaches we are going to conclude that which algorithm is best algorithm and which provides the best result for the model.
