New blog posts
4 March, 2020 by Diana Ferranti
Hello everyone! Here we are today to...
12 February, 2020 by Diana Ferranti
Hello everyone! Welcome back for even...
30 January, 2020 by Diana Ferranti
Hello everyone ! A good night sleep is...
Calendar of Events
Please, configure this plugin in Admin Panel first.
How to build a family
Posted on 4 March, 2020 by Diana Ferranti
Here we are today to understand more about how we constructed our stress families.
As a reminder, once you gather all the data together in the app, from the questionnaire, the sensor device etc… the app returns the stress family to which you belong, as shown below:
Interesting? Sure!! But let us go deeper: how have we found these groups?
We have used a clustering, or data partitioning, algorithm. These algorithms merge a set of data into subgroups (or cluster) sharing common characteristics. In this case, similarity in the way person manage stress and the type of stress that affects most of them.
In our preliminary study, we had a great variety of data (questionnaires, physiological data, continuous measures, one-time measures, etc.). Faced with this large variety, several hierarchical methods were tested in order to find a converging algorithm (unsupervised clustering). The retained method is a Gaussian Mixture Model (or GMM); this model is usually use to estimate the distribution of random variables, as they were Gaussian. For each variable, a Maximum Likelihood Estimation (or MLE) optimizes the mean, variance and amplitude of theses Gaussian.
8 clusters have been retained thanks to this technique (C0-C7 groups).
A second analysis, an agglomerative Hierarchical Clustering Analysis (or HCA), leads us to merge some of these clusters and obtain the 5 large classes of stress, as we know them today.
In the future, these families are meant to evolve! Indeed, they are currently based on an active working population. Our aim is to adapt and integrate all kind of Evimeria® users’ profiles (students, seniors etc…).
However, for now, let us go beyond all these data and algorithms, what are we really talking about, what is « stress »?
See you next time for the answer.
In the meantime, Go Evimeria & Stay tuned!!!