LocationMind: Mobility as a Healthcare

Noncommunicable diseases have become top causes of mortality all over the world with faster aging population structure. However, the prediction for people’s Noncommunicable disease incidence is complicated. So, we develop an APP expected for urban personal health assessment based on their daily trajectory for the improvement on personal health. We have analyzed billions of trajectory data in Japan and Japanese regional health data and found a clear and robust correlation between regional travel mode preference and health level. Then, we can apply this correlation to a deep learning model to assess one’s noncommunicable disease incidence. The input of this model will be his or her personal daily trajectory data. With this technology, we believe in an improvement in global health situation.

The growing trend in mortality population caused by chronic and mental diseases has caused many problems in Japan. Efforts for dampening this trend has become the national level priority of Japan. Thus, a health assessment is required to help improve the national health level. However, how to easily, precisely and generally assess individual health level and predict disease incidence in real time as well as provide health advice with pertinence is widely and urgently required by among governments, scholars and individuals. Recently, many scholars concerned the impacts of regional travel behavior on mortality caused by noncommunicable diseases and injuries. These studies inspired us much on assessment of individual lifestyle and health level by mining his or her travel trajectory data. The main challenge includes the process of raw GPS data collected from individual mobile phone or other portable devices; the deep learning technology to reveal individual lifestyle according to trajectory data; the assessment of individual health level and disease incidence based on massive data statistics and data fusion. Such massive data-based analysis has never been down before. Our work will cover the whole Japan and affect over 100 million people.

Our work will firstly serve for all Japanese residents. By mining massive amounts of trajectory data of all Japanese residents, we discover the relationship between travel habits and some diseases, which is conducive to improve public health. In addition, we have developed many GPS data process technologies and deep learning for revealing individual lifestyle. Based on these technologies, we can develop an application with the IT company and release it to public. Mobile phones or other portable devices with our application will send their users’ daily trajectory data to our server. By analysis, we will send the result of health level assessment periodically to users to remind them of health risks and corresponding health improvement advice.

We will establish a personal health level assessment help analyze the risk of causing Noncommunicable diseases and injuries based on recording and analyzing their daily trajectory data. We will make this function an application installed in users’ phones to record their daily trajectory for the analysis. Then, the application will provide information of the risk for a person of catching Noncommunicable diseases and injuries and provide the corresponding advice on their daily travel mode choice, such as, they should try other travel mode to reduce the possibility of catching some diseases. The main steps are as follows:

• Based on collected massive GPS trajectory data and statistics data on mortality population for different kinds of noncommunicable diseases, we will do comprehensive analysis on the correlation between crowd travel preference and noncommunicable disease. The travel segmentation and travel mode detection will be adopted to extract the travel feature of people’s travel preferences.
• After obtaining this correlation, we will build it to a model to compute the risk of catching main kinds of noncommunicable disease and injury as the input is the personal trajectory data. The advice of changing daily travel behavior to reduce noncommunicable disease will given based on computation result. The analysis and advice will be different based on their location in different areas in Japan. Since features, such as, terrain, weather, in different prefectures can lead to different travel preferences, these preferences can have an impact on regional noncommunicable disease and injury incidence from the aspects of exposing at outdoor, absorbing air pollution, lacking of exercise and so on. The assessment and advice will depend on regions.

We will develop a novel system to finish health level assessment. We will use some trajectory data mining method like advanced travel mode detection, travel segmentation and non-supervision cluster for data processing. The deep learning method will also be adopted to track individual lifestyle based on trajectory data. Many statistics methods help us make a regression on the correlation between individual lifestyle and lifestyle-related disease incidence.

Our application is simple and general. After adapting the correlation to the deep learning assessment model and making it an application in user phone, the utility of this application is very simple. Users just need to carry with them everyday and the assessment can be done automatically based on their daily mobility records.

At last, this proposal bring about a new commercial concept- Internet of Health. With the increasing scale of users, we can collect and cluster their assessment report to build an Internet of Health for future health improvement.

Data Process: Firstly, the travel mode detection, map matching, travel segmentation and inter-prefectural clustering were successively adopted to disaggregate the trips in every prefecture. Then, the statistics method was adopted to count and analyze the travel preference of each prefecture. 

Lifestyle estimation model: We used the LSTM deep learning network to establish lifestyle estimation model based on individual mobility trajectory. This model can effectively estimate individual’s average daily travel time, travel mode, amount of exercise, sitting still time, leisure time, work time, etc.

Correlation analysis and application: We utilized the regression model on individual travel preference and noncommunicable disease incidence and obtained their correlation. Next, we applied it to the deep learning assessment model to enable it to assess the personal noncommunicable disease incidence based daily GPS trajectory record. Finally, it was processed to an application for real use.

After analyzing grand amount of trajectory data and regional health level statistics data, we have established robust correlation between crowd travel behavior and regional health level (noncommunicable disease incidence). Many papers from other scholars have proven the correctness of our work. Based on this, we can provide the service of assessing individual risk of these unhealthy issues. This service will appear in the form of application in users’ phone. It will collect users’ daily GPS trajectory data as the input to output report on the noncommunicable disease incidence. If people can adapt these advices, we believe that an improvement in the individual health can be realized and it can reduce the incidence of noncommunicable disease in a long term.

For the time being, we haven’t provided the service to serve people. But our lab is now cooperating with an IT company in Japan and in one year, we plan to develop a prototype of application in phone with an IT company in Japan and take a test among users. It can be estimated that in one year, 1% users can benefit from this service, which is about 1 million. In 5 years, we plan to extent user scale to 5% - 10%, which is about 5 million to 10 million.

In the future, the main barriers will be in the following aspects:

• Finance: We believe our technology can help more people gain a healthier life. But in the early period, we expect the application is freely open to public to maximize user scale and offer more help. Thus, how to profit in the late stage is one of the main barriers. This barrier requires some funds to group a team to promote the application to some huge IT enterprises with charge or absorb stable advertisement.
• Law: For the time being, there are some laws concerning the utility permission of individual mobility data. How to operate the system precisely with less personal information with the law is another challenge.

In the future, we will overcome these barriers from the following aspects:

• Finance: Currently, we plan to start from two perspectives: The first one is to seek cooperation with some huge IT enterprises. We will pack and sell out our technology in exchange of some funds for our early operation. The second one is the funds from national science research foundations or other company foundations as the supplement of our funds for that our research is attached to The University of Tokyo. We can have access to these foundations.
• Law: we will devote ourselves to developing a precise lifestyle estimation model based on less amount of mobility data to face the data privacy problem with the global promotion of our application and complete of law in the future.

Full-time staff: 4 persons

Part-time staff: 5 persons

Contractor : one person

This solution is based on the existing research results of Intelligent Perception and Urban Computing Laboratory (IPUC), The University of Tokyo. IPUC aims to develop novel algorithms, cutting-edge technologies, and applicable systems to sense human dynamics and mobility, so as to understand human activity and behaviour, and to make cities more desirable, liveble, sustainable and green. Our goal is to improve lifestyle safety, convenience and intelligence, for individuals and the community. We have concentrated on several research themes for about 10 years, such as Emergency Management and Urban Computing, Intelligent Urban Surveillance, Machine Perception, Autonomous Robots, Pervasive Computing Systems and etc.

Our business will start from serving the users of an IT company in Japan as we plan to develop this kind of application with them. In the coming future, we will also cooperate with other companies to obtain global mobility data to establish a global model. Our main income is estimated to be the sale of application to huge IT enterprises, health related company or government. The rest will come from the advertisement in application. 

Now, the main fund source is from LocationMind. The copyright of application and corresponding technology will also belong to this company. This company will provide all the funds to support the development of application. The research fund is from the national science research foundation provided by The University of Tokyo. The current funds are estimated to support the algorithm design, application development and online test for about 3 years. In this period, we will communicate with huge enterprises and government to sell out our application at the purpose of sustainable financial development.

We expect more financial support from Solve and promotion of our influence by win the Prize.

We want to partner with some IT enterprises, health-related company or government who need our technologies or models for further development to more universal business fields. 

The core component of LocationMind is handling and processing big and heterogeneous human mobility data. Comprehensive analysis on these massive data and noncommunicable diseases requires huge computing resources. We are eager to be considered for this prize to upgrade our system to meet the pressure of growing users, develop new architecture APP and website, and collaborate with more data providing companies.

The core component of LocationMind is handling and processing big and heterogeneous human mobility data. Comprehensive analysis on these massive data and noncommunicable diseases requires huge computing resources. We are eager to be considered for this prize to upgrade our system to meet the pressure of growing users, develop new architecture APP and website, and collaborate with more data providing companies.

The core component of LocationMind is handling and processing big and heterogeneous human mobility data. Comprehensive analysis on these massive data and noncommunicable diseases requires huge computing resources. We are eager to be considered for this prize to upgrade our system to meet the pressure of growing users, develop new architecture APP and website, and collaborate with more data providing companies.