Hello Jarvee users and Adnan Cokic (boss of Jarvee),
If you would put all serious Jarvee users together for a whole day to freely discuss best practices, the result would be disappointing. Everybody has its own ways of making it more or less work, but those ways would probably not or partially work for others. Why? Because they have another history, another proxy, another country or other differences in all of those variables which make up a social account.
However there is a solution for a level playing field with the social media, an intelligent solution.
That solution consists of:
1 Periodically collect all relevant data of the social accounts of Jarvee users
2 Analyse them with deep learning techniques from the AI to construct the best model to predict the survival probability of a social account
3 Build that model into Jarvee to predict survival probability of a given set of parameters of a social account
4 Update model
1 Periodically collect all relevant data
Jarvee could make a snapshot of all parameters of every social account every x hours and send it to a server. The same procedure would be followed if an incident occured. The parameters would also include result variables like number of followers, following and general parameters like country. Data collection would only be done with new accounts so that data during the whole lifetime of a social account can be analysed.
2 Analyse them with deep learning techniques / model construction
After some months the dataset would be large enough to start deep learning sessions. This is not very difficult. With tools / languages like Python, Panda, Scikit-Learn, TensorFlow and Kears it is a snap to construct and train models. The resulting model would predict the survival probability of an account given its history and the current set of parameters.
3 Build that model into Jarvee
For any account, Jarvee would calculate in real-time the probability of survival. By modifying certain parameters and watching the survival probability, users can determine a better strategy.
How? Let us say that a Jarvee user sees a survival probability for a given account of 81%. He wants to go faster and increases the number of followers per hour, from 5 to 10. By doing that he sees that the survival probability decreases from 81% to 79% and concludes that is a small price to pay for doubling the speed.
4 Update model
The most difficult part is probably the maintenance of the model. Data collection would have to continue after the first phase to constantly train the model and adapt it to the modifications the social media are working on and implementing.
That means that every installed Jarvee would constantly send snapshots of all accounts to the server. There the snapshots would train the model resulting in new parameters to predict survival probability of a social account, which would be sent to the installed Jarvee’s.
Final thoughts
This is just an example of implementing an AI solution to the survival problem. There are other ways to implement an AI solution.
However there is no no alternative for the underlying basic solution of the problem. That is, continuously collecting data of the social accounts run by Jarvee and analyse them together (as a whole) and doing that with deep learning techniques.
Paul
