Health Monitoring and Disease Prediction System

There has been a tremendous growth in the medical industry over the years, due to the advancement in technology and an increase in health problems have been observed. Due to the hectic and busy schedules of people has led to increased health issues. But as almost everyone in recent times carries their smartphones a healthcare android application can prove very beneficial. The intention behind this project is to create an android application which can be used by people for managing their health. This system will use decision trees algorithm for prediction of diabetes. The user will have to answer a questionnaire which will consist of various parameters regarding the user’s health. This application will consist of features such as Order medicines, Book doctor’s appointment, medication and diet reminders. Decision trees use a tree structure to build the classification models. It divides a data set into smaller subsets. Leaf node represents a decision. Based on feature values of instances, the decision trees classify the instances. Each node represents a feature in an instance in a decision tree which is to be classified, and each branch represents a value. Classification of Instances starts from the root node and sorted based on their feature values. Categorical and numerical data can be handled by decision trees.

Data mining is the method of examining huge pre-existing databases for generation of new information. Diagnosing and Predicting an individual’s health is an important goal of this project. It can be achieved by using advanced machine learning algorithms of the decision tree. An Android application is one of the easiest methods for a person for health management due to the increased use of smartphones. The application will consist of a set of questions which the user will have to answer. This will then use Decision trees which is a data mining algorithm. This will then immediately predict the possibility if diabetes. Besides this, the app will also consist of features like diet monitoring where a user can monitor his/her diet and the app will send notifications or reminders about the same. Other features will be the provision of the book a doctor’s appointment. The app will also allow the user the order medicines. Different factors like gender, age, blood sugar level, cholesterol, hereditary disease and many other factors are taken into consideration in this proposed system. Thus to find out whether the user is prone to diabetes or not.

The main objective is to predict the occurrence of diabetes and monitor a person’s health based on the answers provided by the user of the questionnaire. Other features also include diet monitoring, booking doctor’s appointment, and ordering medicines. There are many predictions algorithms used but because of the parameter consideration or algorithm inefficiency, the accuracy is not so high. Hence, we are considering many parameters and also using C5 Classifier algorithm which gives a high accuracy.

The rest of them into 6 sections. In section 2 and 3, related work and proposed work is presented paper is structured. In section 4, the proposed methodology is presented and details regarding the algorithm to be used are explained. In section 5, the conclusion and finally in section 6 future work has been included.

Health Monitoring Android Application and Diabetes Prediction using Data Mining Techniques [1], in this paper, the authors have proposed a project which seeks to apply information and to create an android application which can be used by patients for management of their health care problems and would thus enable them to have a good life. The applications also create a system for predicting whether a person has a risk of developing the disease of diabetes in the next 10 to 15 years. The system uses questionnaire method using Naive Bayes algorithm A Data Mining Approach for Prediction of Heart Disease Using Neural Networks [2], the authors have proposed a Heart Disease Prediction system (HDPS) is developed using the Neural network. The HDPS system predicts the likelihood of a patient getting a Heart disease. For prediction, the system uses blood pressure, gender, cholesterol-like 14 medical parameters. Here two more parameters are added i.e. smoking and obesity for higher accuracy. From the results, it has been seen that neural network predicts heart disease with nearly 100% accuracy. Disease Predicting System Using Data Mining Techniques [3] proposed a research paper which uses data mining for better disease prediction. It uses Medical data mining techniques like classification, association rule mining, clustering is implemented to evaluate the different kinds of heart-related problems. Survey of Machine Learning Algorithms for Disease Diagnostic [4] provides us the comparative evaluation of different machine learning algorithms for diagnosis of different diseases such as diabetes, heart disease, liver disease, dengue disease and hepatitis disease using medical imaging. It brings attention towards the suite of machine learning algorithms and tools that are used for the analysis of diseases and decision- making process accordingly. C5.0 Algorithm to Improved Decision Tree with Feature Selection and Reduced Error Pruning [5] compared ID3, C4.5, and C5.0 with each other. Among all these classifiers C5.0 gives more accurate and efficient result. This research paper used C5.0 as the base classifier so the proposed system will classify the result set with high accuracy and low memory usage. Feature selection technique assumes that the data contains many redundant features, so it removes those features which provide no useful information in any context. This paper also uses Reduced Error Pruning Technique which is used to solve the overfitting problem of the decision tree.

Decision trees can handle both classification and regression problems. Decision trees require only a table of data with which they will build a classifier directly. On the other hand, Naive Bayes requires you to build a classification by hand. If given a bunch of tabular data, it will fail to pick the best features which can be used to classify. So, for this application, the prediction of diabetes will be done using C5 Classifier Decision Tree.

This application requires the user to answer a questionnaire which will be related to symptoms of the patient and his habits. This will be taken as input and by using decision trees algorithm the possibility of diabetes can be predicted. The application has a feature where the user can enter his/her diet. This app will monitor it and give the required analysis. The medication consumed by the user will also be monitored. The user can also book a doctor’s appointment using this app. The details of the appointment will be sent to the doctor.

The following are the modules of the system:

1. Prediction

2. Diagnosis

3. Diet plan

4. Reminders and alarm and

5. Ordering medicines.

The possibility of the user being prone to diabetes or not will be accurately predicted by this system using prediction algorithm computations. New users need to register themselves in order to create their respective accounts in the system. Access to the system is provided to the user through a login interface. The user can login using username and password on which the user will be directed to the home page of the system.

The user can then enter data for prediction of diabetes. We are mainly focusing on Mellitus type of diabetes. The user can also use the additional features like maintaining a diet, reminding of dosage, call for an appointment of a doctor, etc. The prediction will be done using C5 Classifier Decision Tree. The following are the parameters which would be considered in the data set for the prediction algorithm:

1. Gender

2. Age

3. BMI

4. Blood Pressure

5. Smoking

6. Weekly exercise

7. Consumption of Salty Food

8. Consumption of Alcohol

9. Work Stress

10.Family History of Blood Pressure/Diabetes

11. Pregnant (if female)

12. Blood fat level

13. Health food consumption

14.Fatty food consumption

Decision Trees comes under Supervised Machine Learning (that means we need to explain the input and output in the training data) where continuously splitting of data takes place according to a certain condition/parameter. The tree has two entities, the leaves and decision nodes. The final outcome is in the leaves. And the decision nodes splits the data.

Let us suppose we need to find whether a person is healthy or not healthy. We have parameters like age, type of food intake and exercise time. Here we have a decision either yes or not, this can be called as a binary classification problem. The final output at the leaf would be healthy or not healthy.

There are two main types of decision trees

The above example is of the binary type of classification (Yes/no), where the output is either healthy or not healthy. But here the decision is variable, that is categorical, that is the output is continuous.

Regression trees (continuous data types), Steps for Decision tree:

Algorithm: Generate a decision tree from the training tuples of data partition, D.

Input: Data partition, D, is a set of training tuples and their associated class labels; list of attributes, a set of candidate attributes; Attribute selection method is a procedure to find out whether splitting criterion best partitions the tuples into individual classes. This criterion consists of a splitting attribute and, possibly, either a split-point or splitting subset.

Output: A decision tree.

Algorithm:

(1) create a node N;

(2) if tuple in D is all of the same class, C, then

(3) return N as a leaf node labeled with the class C;

(4) if attribute list is empty then

(5) return N as a leaf node labeled with the majority class in D;

(6) apply Attribute selection method (D, attribute list) to find the “best” splitting criterion;

(7) label node N with a splitting criterion;

(8) if the splitting attribute is discrete-valued and multiway splits allowed then

(9) attribute list attribute list - splitting attribute;

(10) for each outcome “j” of splitting criterion

(11) let Dj be the set of data tuples in D satisfying outcome j;

(12) if Dj is empty then

(13) attach a leaf labeled with the majority class in D to node N;

(14) else attach the node returned by Generate decision tree(Dj, attribute list) to node N; end for return N;

C5 Classifier:

The classifier is tested first to classify unseen data and for this purpose resulting decision tree is used. Each algorithm follows rules or preceding algorithm. Similarly, the C5 algorithm follows the rules of the C4.5 algorithm. The C5 algorithm has many features like:

• The large decision tree can be viewed as a set of rules which is easy to understand.

• Noise and missing data’s acknowledgment is given by C5 algorithm.

• Overfitting and error pruning is solved by the C5 algorithm.

In classification technique, the C5 classifier can differentiate between relevant and non-relevant attributes.

The algorithm of C5 Classifier:

1. To make the tree Create a root node

2. Check the base case

3. With the use of Genetic Search Apply Feature Selection technique best tree = Construct a decision tree using training data

4. Apply Cross validation technique 1. Divide all training data into N disjoint subsets, R = R1, R2,..., RN 2. For each j = 1, ..., N do

5. Test set = Rj

6. Training set = R – Rj

7. Create a decision tree using training set+

8. Decide the performance accuracy Xj with the use of Test set 3. Reckon the N-fold cross-validation technique to estimate the performance = (X1 + X2 + ... + XN)/N

9. Reduced Error Pruning technique is used to find the attribute with the highest info gain (A_Best) Classification: For each tj D, apply the DT to determine its class.

This report provides the comparison of a different system for disease prediction. We have studies various data mining and machine learning algorithms and have come to the consensus that Decision trees are the most efficient. The algorithm C5 classifier, which we are using predicts disease to a better extent. The application is software specific, hence it is cost effective and can be widely used. The application provides a diabetes prediction system which would help the patients take precautions and hence avoid or prevent the occurrence of diabetes. The automated message facility for major fluctuations of vital signs can be useful during the time of emergencies. The application also helps in regulating the diet of a person and also the reminds him/her of medicines dosage. Hence the proposed system will make the process of healthcare management quite easy and efficient.

The current system only deals with the prediction of diabetes. Using advanced machine learning algorithms and data mining the system can be updated for other diseases. The system can be updated to take the patient’s blood report as an input and determine the possible diseases or complications that may arise. In the future, the concept of image processing can play an important role in the prediction of diseases, the patient’s health can be predicted using images of the patient. There are also many types of diabetes, as we have selected mellitus diabetes, furthermore, applications can perform prediction of other types of diseases using machine learning algorithm or image processing. The application in near future can search hospitals according to the area in which the user is located. It is impossible to generate or get a dataset of all the existing individuals, but it can happen that someone generates data set with many parameters and data of almost every individual.