Page 249 - Big Data Analytics for Intelligent Healthcare Management

242 CHAPTER 9 INTELLIGENCE-BASED HEALTH RECOMMENDATION SYSTEM

The presence of outliers and inaccurate records or data will affect the whole analytic process. Be-
fore starting the data analytic process, it is necessary to clean the data. In this phase, corrupted data or
missing data generated during data collection is elucidated by replacing the missing data with an av-
erage, mean, or global variable. Noisy and irrelevant data are removed during this process. The next
stage is known as data analysis. This part of the framework is responsible for processing different data
types and implementing analysis of data. In this part, the process of data analysis can be categorized
into three main techniques on the basis of data type and the purpose of the analysis.
MapReduce is a widely used, general purpose computing model and runtime system for distributed
data analytics, which has the capability to process huge quantities of data efficiently and also analyze
huge quantities of unstructured data concurrently in a multiprocessing structure. The MapReduce tech-
nique has been used to enhance the rapidness of medical image processing, finding optimal parameters
for feature classification by applying the machine learning technique, data mining, SVM, etc. Stream
computing enables hospitals to process data streams in near real time. Stream computing helps system
to spot opportunities and risks across all data. Data analytics is the process of performing analytics
directly on the database where data resides, therefore, eliminating the need to move data. The infor-
mation and/or ratings of the doctors can be sent to end users through the internet. Sometimes rating can
be sparse or inaccurate due to patient’s reluctance and confidentiality of medical data. If the patient,
who is participating in the process, is assured about his privacy then trustworthiness in the system in-
creases and people working in various hospitals, clinics, and dispensaries can work in cooperation with
authorities in a secure manner. As a result, appropriate and precise predictions can be produced.
The system reads the symptoms of patients and identifies the disease from the database and recom-
mends disease-specific doctors to patients. The sensors, medical instruments, and recording devices
record the symptoms of the patients such as high cholesterol content, low blood sugar, headache, in-
digestion, and high body temperature, and then transfers the patient’s information to a smartphone via a
centralized server. Moreover, the information can be sent to the recommendation system via a tablet,
laptop, or wireless network. The HRS then suggests the name of a specialist to the patient for treatment.
Here collaborative filtering (CF) is used as the filtering technique to process large quantities of health-
care data due to higher accuracy. CF recommendation engines are more versatile, in the sense that they
can be applied to any domain, and with some care could also provide cross-domain recommendations.
The overview of the health monitoring scenario using HRS architecture is shown in Fig. 9.6. The
feedback or ratings about doctors is transmitted by patients after receiving the treatments via the HRS.

9.4.1 DATASET DESCRIPTION
We used a healthcare dataset on the proposed intelligent HRS. This healthcare dataset contains discrete
ratings from 1 to 5 of 10,000 patients for 500 doctors. This dataset was divided into training and test
data in 75:25 ratios respectively. A 10-fold cross-validation scheme was used to evaluate the results.
Our proposed HRS was designed and tested on healthcare dataset which describes rating information
along with further details. The experimental results are described in Section 9.4.2.

9.4.2 EXPERIMENTAL RESULT ANALYSIS
Here we compare the results in terms of MAE value among existing methods and the proposed HRS by
analyzing the healthcare dataset. As we obtained a lower MAE value for our proposed approach, we can
say that our approach is a useful healthcare recommendation system.