Early Cancer Prediction and Hospital Recommendation using Integrated Fact Based Information and Opinion Summarization with the application of Knowledge Based System and Data Mining Techniques

Abstract

In this digital world, the support of health care system recently with the help of computer application are tremendously changing the way health care organization are currently delivering their services to the users. Recommender systems (RS) suggest items of interest to users of information systems or e-business systems and have evolved in recent decades. A typical and well known example is Amazons suggest service for products. Patient disease prediction, advising and suitable health care recommendations are some of the services that are provided in traditional means in hospitals that are found in Ethiopia.One of the problem currently facing healthcare/hospital organization is the provision of quality services at a reasonable price. Quality services can be related to customer satisfaction, correct diagnosis, and proper management of patient treatment in Developing countries like Ethiopia. This is due to absence of serious preventative mechanism that can be followed. The main objective of this study was to develop cancer disease prediction and recommendation Model at Early Stage using integrated fact based information and opinion summarization with data mining and knowledge based system. For prediction of cervical cancer J48 Algorithm was selected over 10127 records of cervical cancer medical history. Hand crafted rule with Vector Space Model was used for opinion summary. For fact based information only Vector Space Model was used. Afaan oromoo and Amharic dataset were collected in two versions: the first version is for training set and the seconed version is for testing set.In one hand, the dataset for training was Prepared in two forms:the first one is the data from online English patient opinion is translated into Afaan oromoo and Amharic using English to Afaan oromoo and Amharic language translators. The translated opinion was provided for Afaan oromoo and Amharic linguistic experts to correct its language meaning and the polarity of the sentiments. Then the corrected opinions were used to train the model with selected hybrid approach.An experiment was conducted on five hospital in Addis Ababa: ALERT, Black Lion Hospital, Teklehaimanot General Hospital, Zewditu Memorial Hospital, and St.Pauls Hospital. Lastly, as we understand from an experiment carried out in Amharic and Afaan oromo version Reviews St.Pauls hospitals registered less score in every experiment.Therefore, such types of output can’t be considered as patient language complexity problem but there is some fundamental problem that should be addressed by the health care. For integration of data mining with knowledge based system JPL (java prolog language) was employed for integration. For this, User acceptance testing was carried out and registered 90.6 % of accuracy that was promising result. The developed model was developed and applied on five hospitals data set. The recommender system is scalable and works for more than one disease if the EDPHR Model are supplied with adequate and proper data. The developed EDPHR Model can also help both patient who already know the disease and the new patients .for the patient who already know the disease it will recommend the best choice hospitals. for new patient it will detect the disease at its early stage and recommend the best choice among the hospitals. The developed system can minimize the disproportion between the patients and professionals.Generally, efficiency of opinion summarization depend on the frequency of summarization. The more input is summarized, the more efficient and effective the algorithm will be. The future direction has been created for The developed Model can works only for medical domain, but in the future we planned to extend the system to develop model for other domain such as university, hotels and banks