A Digital Health Companion, Virtual Banking Assistant & Smart Water Monitoring System – All Built on Microsoft AI

07

Sep

A Digital Health Companion, Virtual Banking Assistant & Smart Water Monitoring System – All Built on Microsoft AI

A Digital Health Companion, Virtual Banking Assistant & Smart Water Monitoring System – All Built on Microsoft AI

Source: https://blogs.technet.microsoft.com/machinelearning/2017/09/07/a-digital-health-companion-virtual-banking-assistant-and-smart-water-monitoring-system-all-built-on-microsoft-ai/

 

 

 

A selective peek at new customers benefiting from Microsoft AI & Machine Learning.

Ruppiner Kliniken Creates a Digital Companion That Can Save Lives

17 million people die of cardiovascular disease each year. Many of these victims could survive if their physicians became aware of critical changes occurring in their vascular system and heart in a timely manner. Unfortunately, in many such cases, there are no external symptoms to report. In other cases, patients fail to report symptoms to their physician until it’s too late.

The cardiology department at Ruppiner Kliniken, a leading hospital in Germany, was motivated to save the lives of as many such patients as they could. They realized that they needed a way to collect patients’ data health in an unobtrusive way, without interrupting their daily lives. When fitness bands hit the mainstream market a few years ago, they realized their potential in improving patient care. By enhancing the fitness band with new types of tiny wireless sensors that collect ECG, blood pressure, and a variety of other medical readings, it would be possible to collected a patient’s data over weeks, months, or even longer, and consolidate and analyze such data to get a much more accurate view of their health.

The hospital partnered with a consortium of 37 organizations – including healthcare and medical devices companies, IT startups, and universities – to create an IoT proof-of-concept called Digital and Analog Companions for an Aging Population (digilog). The system captures and sends a patient’s ECG readings and other health measurements to Microsoft Azure. There, IoT (Internet of Things) and other services transform disparate medical data into easy-to-read summative dashboards and key performance indicators (KPIs), which patients and physicians can view from a mobile app or web portal. They can see high-level and drilldown views of how a patient’s heart and cardiovascular system is operating, both when the patient is at rest and when active. They will also see predictive models to help guide what treatment or lifestyle changes may be necessary.


The consortium chose Microsoft Azure as their IoT platform of choice because of its scale, low cost and ability to work with virtually any device. The system uses Azure IoT Hub to ingest data from devices in real time. Information on each patient is stored separately and protected against unauthorized access with Azure security services. Automated algorithms developed with Azure Machine Learning intelligently analyze each patient’s data over time.

The creators of digilog are testing the solution and resulting data extensively for accuracy, reliability, and security. They are also evaluating the feasibility of giving people the authority to own and control their medical information and associated alerts. For instance, they can configure alerts to be sent to specific persons when certain conditions occur – especially useful for patients who are impaired or live alone. They also hope to make healthcare more affordable and accessible through telemedicine programs that reach rural and underdeveloped populations.

Ruppiner Kliniken and their partners are excited by the potential of digilog to reduce risk, save lives, lower healthcare costs and improve outcomes for more people, regardless of where they are, by weaving medical care into people’s everyday lives.

Read more about the Ruppiner Kliniken story here.

Nedbank Introduces EVA, the Electronic Virtual Assistant

Serving individual investors directly, via rich one-on-one interaction. is a top priority at Nedbank, one of South Africa’s biggest banks that operates out of seven African countries. Such interactions mostly happen through their call center today, as that’s what customers find the most convenient. The challenge for the bank has been to replicate the convenience of the call center interaction but in a more cost-effective channel, one that’s just as readily accessible by their clients.

Increasingly, that preferred alternative channel is the messaging app. Some of the most popular apps in South Africa (and in much of the world) are WhatsApp, Facebook Messenger, Slack and their ilk. Most of the bank’s clients already use one or more of these apps. Thus, the bank started exploring cloud-based chatbot technology to scale their client service in a cost-effective manner.

When evaluating a range of popular bot platforms, Nedbank found some too inflexible, others too expensive, or too basic for their needs. They finally decided to go with Microsoft’s Bot Framework and LUIS (Language Understanding Intelligent Service) as it gave them the power and flexibility they needed.

Nedbank named their proof of concept EVA – short for the Electronic Virtual Assistant. Their immediate goal was to test client interactions with EVA for their top 10 most common inquiries. They got the EVA prototype up and running in just three months, with a more feature-rich version went into production less than four months later. That’s very fast for the world of complex financial services, especially given that these were being placed into production usage. Nedbank credits the easy-to-use technology in the Bot Framework, its connectivity with downstream messaging apps, and the instantaneous on-demand commissioning and decommissioning of infrastructure as being key factors in the speed of their deployment.


Most of the development cycle had to do with linguistic rather than technology considerations. The team had to work with linguistic experts to help EVA identify the intent of questions, to ensure appropriate responses. Such localization is critical to the success of a chatbot service. For instance, the answer to “Which is your highest-performing fund?” is different depending on whether the client’s previous question concerned equities or bonds. Successfully meeting those challenges required an iterative process that included people outside of software development, such as from their investment marketing department.

EVA went live in February 2017 and was so successful that many clients thought they were interacting with live agents. The bank observed that EVA could handle 80 percent of the inquiries for which it’s programmed but at just 10 percent of the cost. Even better, EVA frees up the bank’s human agents to handle the more challenging and tricky situations that need live agents, rather than handling the more routine or mundane questions.

Clients access EVA through the bank’s investments website, but the plan is to make EVA available through messaging apps, fulfilling Nedbank’s vision to meet clients via channels they already use. Nedbank also plans to expand EVA to assist clients with transactions, and are also evaluating applications in other areas – for instance in insurance, vehicle financing, and business and retail banking.

You can learn more about the Nedbank story here.

Powel Drives Smarter Water Monitoring and Distribution

How does one make water smarter?

Municipalities in Norway see an average water loss of over 30 percent due to leakages in their distribution networks. This high rate of leakage has both a financial and an ecological cost, and a key contributor to the problem is the age of the components in their distribution infrastructure.

Developers at Powel, a Norwegian software company, through their careful collection of sensor data, have been able to successfully transform currents of water into streams of data. They analyze these data streams using machine learning algorithms to locate problems in municipal water supplies and alert the appropriate utilities, saving both money and a vital natural resource.


Powel’s software developers faced two key challenges: First, to determine what a normal water flow looks like, especially when there are already leaks in the system; next, to efficiently monitor water flow over time, to detect new leaks.

The developers realized they already had some data to work with. In most municipalities, water flow is tracked using supervisory control and data acquisition (SCADA) sensors. Using the municipality of Trondheim as a test case, they retrieved SCADA data between 2013 and 2015. They used that data to build an Azure Machine Learning model that could be used to predict normal water flow throughout the year. For the ML algorithms to be useful in determining abnormal flow due to leakage, the results predicted by Powel’s engineers needed to be compared against live data from the SCADA sensors. Team members quickly settled on a tool for ingesting real-time sensor data for analysis. Given the sheer number of sensors and the volume of data, as well as their security and device management needs, they decided to go with Azure, including the Azure IoT Hub, as their solution.

During the training of the ML model, the team needed a way to understand the data they were working with. They found Power BI Desktop super useful to explore and visualize their data, allowing them to quickly look for patterns and insights. Using Power BI in this way, the developers identified the fields they needed to include in the dataset and what kind of cleaning needed to be done before experimenting with the data using machine learning.

Through their Power BI pre-analysis, the Powel team learned that the water flow into the water distribution system follows a daily pattern, with some exceptions during special events and public holidays. To allow for a granular model for water flow within a 24-hour period, they split every 24 hours into 5-minute intervals. They also included a flag to indicate whether a given sample time was associated with a public holiday. By combining a machine learning solution for predictive results and an Azure IoT Hub solution for real-time data collection, Powel found a way to make it easy for the municipal water systems of Norway to self-report on their status and even tell the utilities when that status is abnormal.

Powel has effectively made monitoring of the water distribution smarter. The developers who built this project are currently integrating their Water Alert solution into Powel’s main product line. As water utilities install more sensors and smart water meters over time, the solution will have the potential to provide even more intelligence, as real-time leakage warnings are cross-referenced with historical repair data to isolate the sources of leaks.

Learn more about Powel’s Water Alert solution here – you can even review code samples and architectural diagrams created by Powel’s developers.

ML Blog Team

Leave a Reply