Role of Feature Engineering and Classifier Selection for Machine Learning Predictions

Features play a crucial role in several computational tasks. Feature values are input to machine learning algorithms for the prediction. The prediction accuracy depends on various factors such as selection of dataset, features and machine learning classifiers. Various feature selection and reduction approaches are experimented with to obtain better accuracies and reduce the computational overheads. Feature engineering is designing new features suitable for a specific task with the help of domain knowledge. The challenges in feature engineering are presented for the computational music domain as a case study. The experiments are performed with different combinations of feature sets and machine learning classifiers to test the accuracy of the proposed model. Music emotion recognition is used as a case study for the experimentation. Experimental results for the task of music emotion recognition provide insights into the role of features and classifiers in prediction accuracy. Different machine learning classifiers provided varied results, and the choice of a classifier is also an important decision to be made in the proposed model. The engineered features designed with the help of domain experts improved the results. It emphasizes the need for feature engineering for different domains for prediction accuracy improvement. Approaches to design an optimized model with the appropriate feature set and classifier for machine learning tasks are presented.

Machine Learning: A Comprehensive Survey on Existing Algorithms

Machine learning is a subject that reviews how to utilize PCs to reenact human learning exercises, and to think about self-change techniques for PCs that to get new information and new abilities, distinguish existing learning, and persistently enhance the execution and accomplishment. Contrasted with human learning, machine learning adapts speedier, the collection of information is more encourage the consequences of learning spread simpler. In this way, any advance of human in the field of machine learning will improve the ability of a computer, along these lines affect human culture. This paper introduces machine learning, its basic model, and various techniques. This paper also discusses various researches done in previous years in the field of machine learning.

Multicriteria Based Recommender System for Enhancing Customer Centric Business

A multi-criteria-based recommender system makes precise decisions by probing customers multiple criteria’s on-hand and gives recommendations by modelling a user’s utility for the alternatives along several criteria. This paper collects user preferences of criteria over alternatives in the form of linguistic variables. Alternatives includes 6 policies namely term plan, endowment plan, child plan, life insurance plan, criteria include income tax benefit, sum assured, benefits on death and riders option. To rank such alternatives, fuzzy vikor expanded form of MCDM (MCDM) technique is used. MCDM ranks policies based on expand S (S) and expand R (R) value. Sensitivity analysis is used to determine the stability in the alternatives ranking, with the varying parameter value V. The proposed approach is compared with fuzzy topsis approach which ranks alternatives based on closeness coefficient value obtained. After ordering the alternatives using the MCDM techniques, it is inferred that, Both the approach almost provides the closest ranking order. The proposed fuzzy vikor provides best and optimal solution preserving the consistency compared to fuzzy topsis.

Fake News Detection Using Unsupervised and Deep Learning Algorithms

A detection of fake news is difficult due to limited publicly available resources (Datasets). Fake news is a false information which present in news or stories, blog so on. Fake news easily spread and damage the reputation of person or an organisation, therefore, detection of fake news is important. This project work detects fake news using unsupervised and deep learning algorithms. In unsupervised learning method One Class SVM (Support Vector Machine) and in deep learning method Hybrid CNN-RNN is implemented. Experimental results with NEWS dataset showed an accuracy of 58% for One Class SVM and 96.4% for Hybrid CNN-RNN. The proposed method performs better in terms of application performance compared to already existing Machine learning algorithms. This project can be further extended by exploiting high dimensional datasets in future.

Organic Food Store Management System: A Desktop Application for Small Scale Business

Organic food store management system is a desktop application to empower people or a small-scale businessman to manage the store, in each and every way possible. This management system aims to simplify and modernize the work efficiency of the businessman. Organic food store management system is a software developed for the shop where they sell or make only organic food materials which further aims to keep people free form chemicalized food material and helps to increase social as well as geological health. It ethically aims to save mother earth by saving paper, plastic etc. And hence contributes to the good health of mother earth. As we all know that the world and business culture are rapidly increasing, we have to be growth full towards the technology and should be more aimful to save manpower or more handwork towards a very ease technological or computerized shop or business which can further aim in ethical handling of the business. Here, in this software there are three main role players; Admin, Cashier and Customer. Particularly each role player has a privilege to comprehensively work on the software. Various operations such as Bill Creation, Stock Management, Virtual Display of product Info, Display of Pamphlets Virtually and many such, can be done with the use of this software. The genesis of the review pampers to save time, money as well as pamper Mother Earth.