Cigarette smoking is responsible for more than 480,000 deaths per year only in the United States, including more than 41,000 resulting from secondhand smoke exposure.
When was the last time you think to quit smoking? No, I’m not here to give any lecture on smoking but to discuss a pretty beautiful application of Artificial Intelligence that will people to quit smoking.
So, the basic requirements here be will a smartwatch and an app called Cue inside it.
This app works on the fundamentals of habit-forming and has been designed in such a way that it cannot make the user do more work and also deliver a benefit beyond the novelty phase.
Cue is pretty cool and smart, it has an inbuilt algorithm that automatically detects when you smoke, so it keeps counts of your cigarette. The company behind this app said that it pushes users to make small but consistent improvements.
Its machine learning algorithm has been trained on the behavior of smokers habits, like how frequently they smoke, from what point there is an increase in smoking, usual time of smoking in order to understand users current stage of smoking.
Further, it has been also trained over the data of people who actually quit smoking and with the researcher, a plan has been designed that follows a process of a number of steps that can be applied to the user to help him to quit smoking.
After this lots of training, when the users start using Cue, the machine learning algorithm of Cue starts to understand users smoking habit, like if the user smoke, please sign up or otherwise share it with your friends.
The app quickly identifying users with unique movement patterns & discerning smoking from all other actions and start predicting when the user is going to smoke based on individual and aggregate data.
The algorithm that automatically detects when a user is smoking uses a kiwi toolkit in with motion recognition, much like how steps are counted, Cue detects hand to mouth movements to define when a person is smoking.
Another amazing thing about this app is that it’s all our classification algorithms run on the watch, so once the app is download no connection to your phone or internet is needed to make it work.
A neural network is used to separate movements that may overlap such as lifting objects, walking with your hand raised or talking enthusiastically.
Predicting When a Person is Likely to Smoke
In order to best help people nudge the time between cigarettes to longer and longer periods; a prediction needs to make for when your anticipated next cigarette will be, from here we nudge slightly to increase the time by a little bit to keep the effects of the behavior change subtle
Here is a plot of daily occurrence of when people smoke over a week:
Prediction and Why This Is Valuable
Similar to being able to automatically detect when a person is smoking, a method can be applied to predict when a user is smoking in the day, like most model problems the challenge is how to get clean data and ensure you have a benchmark to correlate with.
For instance, a Convolutional Neural Network can be used for photos, where different RGB signals can be separated to define the difference between labeled cat and dog photos.
How can this be applied to just a few data points in a day such as smoking times?
What we do is generate features to generalize across our entire user base to find distinguishing pieces of information, the plot above shows a sample of three features:
Feature 1: the time between current and last cig
Feature 2: current day of week, format: [0..6] -> Mon..Sun
Feature 3: current time of day, format: HHMM
Using these features we are able to come up with a prediction to the nearest 30 mins or 60 mins; with this we can help users pre-empt the urge of when to smoke next by going for a walk, playing a game or grabbing a coffee on us, more to come.