Chemin du Pré-Fleuri 3, CH-1228 Plan-les-ouates +41 78 930 26 37
Machine Learning

For highly evolving system or with a lot of sensors, it is complex to create specific alarms for each cases.

This is why we developed Machine Learning algorithms to enable automatic issue tracking.

Why machine learning ?

Ebeewan use machine learning algorithm to create automatic actions that nowadays would need human intervention.
Predictive Maintenance
Be able to predict when a machine needs to be reviewed will  lead to significant savings.
Anomaly detection
Every issues cannot be anticipated, our algorithms are able to find out inconsistent data.
Performance optimization
Looking for heat loss, energy consumption analyse.

  • Text Hover

Algorithms key numbers

Since their deployment our algorithms have been tested cautiously. From these test we release statistics about efficiency of our machine leanings design.
Accuracy
0%
Error Average
0%
Issues detected
0%

Prediction process example

This process is used to detect, for a given time, 
if the received datas are consistent
1

Simple data model design

Model creation using previous data or similar sensors models already existing .

2

Next data prediction

We use our computed model to predict the next data message.

3

Compute average difference

We compute the average difference that this sensor is used to send (e. g: + or - 2°C)

4

Data controls

At this step we liken received data with the computed one reflecting the data average.

5

Throw alert

If the previous step show an inconsistent data, an alert is thrown.

Popular algorithms used

Our machine learning algorithms are based on famous artificial intelligence process
that we used for our clients.

K-Nearest Neighbour

A light weight computing resources way to find out if a received data is consistent regarding an cluster of data.

Feedforward Neural Network

These heavies algorithms are slow, need to be trained and a lot of computing resources but could discover quality of data and data schema with a great efficiency.

K-Means Clustering

K-Means clustering are used to compare data segment instead of isolate data to look for global sensors behaviours.