Our project is a doctor recommendation system. In this project we aim to provide an online health service to people to make health services easy, effective and better for them. To find a specialized, diagnosis correct, skill superior, outstanding reputation, practical experience, fee economical and at the same time easily accessible doctor is very important and needed for patients but it is not an easy job. Today most recommendation applications are not as much intelligent and up to date as needed for users to suggest and help them in finding appropriate doctor that matches with the user requirement, because inadequacy of doctor’s profiles, multiplicity of user symptoms, patient’s medical history and information mismatch have a great impact to make it difficult for such recommendation applications. To make a personalized recommendation application for providing useful and effective medical services we need user reviews from online communities and doctor’s up to date information from emerging medical databases. We describe an integrated recommendation application in this project to find the best, specialized, good reputation and nearest doctor by considering the user requirements including their illness symptoms, accessibility and the fee. Three models are proposed for this recommendation application. First model is the matching model which suggests the matching between user’s requirement and doctor’s profiles intelligently and accurately. Second model is the quality model which is used to measure the doctor speciality, experiences and consider the user’s opinions, ratings and reviews for doctors to help new users in finding a doctor. This model is build by doing the sentiment analysis of comments about doctors to generate a new rating as well as considering the rating given by previous users. Third model is the finding model which uses GPS service to find the nearest and easily accessible doctor. This model is build by calculating the distance between the user location and the doctor location. Finally, a mobile based application is developed to demonstrate the functionality of the results of these three models.

Abstract

Stock market is a risky environment where many factors can directly or indirectly impact on stock index. These factors are mainly responsible for the fluctuation of stock prices such as industry performance, investor’s sentiments, and economic factors to name a few. This project is intended to provide an awareness to customer for stock predictions of certain companies. Stocks are selected and evaluated at day end and then machine-learning techniques/models are applied to predict whether it is recommended to invest or not. The objective of this project is to generate a recommendation system for small investors and to generate the alert for the large investors in case of any uphill’s/downhills in the prices of the market. Accuracy of results is the most important factor in the prediction systems. Prediction in stock market is very challenging as there is no defined rule to predict/estimate the current market trend. Because of this limitation, many existing methodologies such as text analysis, statistical modeling, fundamental analysis, multiple linear regression will neither work accurately nor produce acceptable results. Although aforementioned methodologies are functioning well in dealing with small data but their limitation is they are unable to deal with large data stored in stock industry. Due to these limitations, the methodology used to create the prediction system is artificially created neural networks. Significant features of stock market that have largest impact on stock prices are selected, then extract their features from big data to predict the fluctuations through neural network algorithm “backward propagation method” to achieve accuracy. The importance of using neural networks are good in dealing with sequences and time intervals and can make good predictions on the data like weather, stock return prices. The main objective of this project is to predict the currencies entails cryptocurrency, oil prices, gold, and some large companies of U.S stock market. These critical share markets cannot be predicted with 100% accuracy because of indefinite real-time constraints, but our designed models can be trained which give a good idea to user for investment time, holding or jump out time of investments.