Artificial Intelligence Project – Speech

Opening Comments:

This project is basically an analysis to probe into the real world implementation of artificial intelligence and understanding how such systems can be potentiated to impact the everyday life of people around the world keeping in view the latest advancements of the virtual infrastructure available.

In that regard I have first thoroughly investigated the need for the intelligent health monitoring systems which can employ artificial intelligence to impact the lives of many people around the globe.

Although the realization of the complete end to end system is beyond the scope of this study and my academic level, but this project has given me a purpose fueled with an idea of creating actual impact in lives of many through the artificial intelligence capabilities.

Along the way I have learned many new things and this project has become a personal passion which I will pursue in the years to come.


So firstly I studied the capabilities of the cloud computing infrastructure which can be employed in a health monitoring system. The concept of the proposed system in its entirety is very huge so I focused on demonstrating the strategies for deployment rather than creating the complete prototype of the cloud based system. I also learned the ways to use different cloud based modules and layers of a system in integration and focused on the predictive analytics layer of the cloud based architecture.

The Main objectives are the following:

To understand the application of machine learning and predictive analytics in cloud computing systems in order to constitute the and realize an end to end system architecture of the Intelligent Health Monitoring system.

To specify the requirements and challenges for building the Intelligent health monitoring system for the consumer market keeping in view the patient and doctor’s requirements.

To implement algorithm for the predictive analytics module of the system, using Cloud SaaS platform.

To identify data handling issues in the workings of the system.



An overview of the MS Azure architecture and layers conceptualized for the system:

Microsoft Azure provides a comprehensive platform to cater all sorts of needs of the proposed systems through the capabilities of cloud computing. The platform embodies the remote sensing infrastructure and modules are available both as a Software as a Service (SaaS) and Platform as a Service (PaaS) for the complete end to end deployment of the proposed system. Given that in this study my focus was on implementing the machine learning algorithm to make the system artificial intelligence, I focused on the “Real-Time Analysis” layer of the cloud architecture of the system.

However, given that one of the objectives of the study was to conceptualize the system implementation in its entirety so I used content analysis technique to understand and construct the architecture of the system.

The system architecture is based on the reference architecture provided in the figure below:

Main concern for the health monitoring system is to define the analytics layer for the system which shall be capable of streaming the live data and applying machine learning algorithms using third party data sources for enabling the system for artificial intelligence in disease discovery and prevention diagnosis. For this system the most appropriate module for streaming and analyzing the data is the Azure Stream Analytics. After streaming the data the system utilizes the MS macine learning studio for the implementation of the machine learning algorithm and the newly structured data is managed and stored using a combination of Azure Blob storage and SQL database.

Device Connectivity: In the real world implementation of the system, the data is acquired through the remote sensing devices attached to the patients in their homes such as the hemodynamic sensor for heart rate monitoring. All the devices are connected to the cloud via a control unit for communication and data transfer. These devices are low power and high power IoT devices connected to the data processing section of the architecture through the cloud IoT gateway which recognizes and streamlines data coming from different devices. The cloud gateway further facilitates bidirectional communication between the system and the remote control units. It generally manages all aspects of communication, including transport-protocol-level connection management, protection of the communication path, device authentication, and authorization toward the system. It enforces connection and throughput quotas, and collects data used for billing, diagnostics, and other monitoring tasks.

After Cloud gateway and provisioning of API the data is transferred to the device identity store which is the authority for all device identity information. It also stores and allows for validation of cryptographic secrets for the purposes of device client authentication.

After the data is acquired through the devices the following procedures take place:

Storage: SQL Database provides both a storage capability as well as a query engine for returning records to a user. Also, Event Hubs provides storage for a pre-defined period of time. This diagram is intended to illustrate each of the services being mapped to their primary role within this architecture.

Analytics and Machine Learning:

Azure Machine Learning provides a cloud based platform for mining data to identify trends and patterns across large scale information sets. By “learning” from historical events and trends, machine learning can publish a model that can be leveraged for determining predictions in real-time based on incoming event data. Combining the real-time rules based processing capabilities of Azure Stream Analytics with the real-time predictive analytics capabilities of Azure Machine Learning can help businesses rapidly deploy highly scalable data solutions to support complex information challenges.
















For the purpose of this study I have implemented algorithms for the machine learning and analytics part of the system. For the experimentation and implementation of machine learning algorithms I have used the SaaS platform provided as a backend.

The implementation is a module from the complete system architecture as demonstrated before. The data which will be accessed in actual system through remote sensing IoT device, for this algorithm implementation is utilized from an external source and is loaded into the MS Machine Learning Studio.

In this experiment, I used the training Heart Disease data set from the UCI Machine Learning Repository. The data was collected from several locations (Cleveland Clinic Foundation, Hungarian Institute of Cardiology, V.A. Medical Center, University Hospital Zurich).

We built the heart disease classification model using data from the Cleveland Clinic ( The training data set used in this experiment consist of the following features:


  • age
  • sex
  • chestpaintype
  • resting_blood_pressure
  • serum_cholestrol
  • fasting_blood_sugar
  • resting_ecg
  • max_heart_rate
  • exercise_induced_angina
  • st_depression_induced_by_exercise
  • slope_of_peak_exercise
  • number_of_major_vessel
  • thal (results from a thallium heart scan)


The experiment can be accessed through this link:


The models are trained in four different sections using the “Train Model” module of the Microsoft Azure Machine learning studio. Then the patients IDs are used to test the trained model. The result output is executed using the following R script: Given in the appendix 1.


Training and evaluation of regression models is then conducted to check which regression model is most appropriate for the acquired data set. The following four regression models are used to train the data: Decision Forest regression, Boosted Decision tree regression, Poisson Regression, and Neural network regression. The data is basically assessed and trained using the following steps:

Classification: The purpose of the Classification model is to determine a label or category  it is either one thing or another. We train the model using a set of labeled data.

Clustering: We would create a Clustering model if we had a whole bunch of data but we didn’t have a determined outcome, we just wanted to see if there were any distinctive patterns.

Regression: A Regression model is created when we want to find out a number – for example how many days before a patient discharged from hospital with a chronic condition such as diabetes will return.

Deep Learning: Deep learning is a buzzword we hear a lot but is often misused, in reality it is just a special case of one machine learning algorithm: artificial neural networks.



Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s