A supervised machine learning algorithm that uses multiple sensors to identify a child's feelings and determine whether their behavior is normal or indicative of trouble.
Is a child just happily exploring or are they destroying stuff as a way of protest? Does a teenager need help or is the extreme new styling just part of a journey of self-discovery?
Kids are not good at putting their feelings into words and thereby seeking the right solutions for their problems. This would assist the caregivers in identifying problems that need to be addressed.
There comes a time when teenagers begin to assert their independence and freedom. They become secretive and it gets difficult for parents to identify whether their child needs help or more freedom.
The goal of machine learning (ML) is to learn from data and make accurate predictions, without being explicitly programmed. The idea behind supervised ML is that you specify a set of input parameters and a result you expect to get. Thus, you teach the algorithm to provide correct answers.
Parents know their children pretty well. They would thus be suited to teach the ML algorithm when their child is happy, feeling satisfied, sad, angry, etc. The child would be monitored by various sensors several hours per day. The parent/caregiver would specify on the child's timeline the exact timeframe when the child felt specific emotions. Based on all the available sensor data at those moments, the algorithm would learn to identify a child's feelings.
The parent could thereafter see a timeline/log of the child's feelings whenever the child is in range of the sensors. They could correlate the feelings to what their child is doing and thereby get a better sense of the child's healthy development.
There are many existing wearable devices (rings, watches, mobile phones, beds) that are full of sensors. They monitor heart rate variability, breathing frequency, pulse, body temperature, movement, etc. Tesla's state of the art self-driving ML algorithm learns from video/audio.
The proposed ML algorithm will make use of any data it can get. The more data is provided, the higher the chances for accurate predictions.
The sensors should be unobtrusive and acceptable to the child being monitored.
What parameters would be useful to measure in this case and how?