How data is analyzed?
The main goals for data analysis are:
- Track errors in your plugin settings
- Track conflicts or duplcated settings if occured
- Check if everything is in working order
- Create spilt tests and analyze the best usage
- Analyze if there are any other chances to achieve your goals
The complete process is based not only on machine learning algorithms, but static analysis too.
When your statistical data comes into our servers, we process it using common algorithms to find any errors or conflicts, suggest any solutions, best usage methods and help you to discover all the functions of the plugin.
For some goals is useful to use static functions to make comparison between your settings and find any conflicts that may occur. For the prediction models and for the suggestions, comes machine learning classification algorithms.
The best usage for tracking errors is statical analysis. This is a set of functions, which are processing your data to find some errors or conflicts.
Each LlamasApps plugin have it's own static functions, based on the specific plugin features.
As an example, let's focus on SEO redirects plugin:
In this case, such conflict might be a duplicated redirection rule. Or a redirect rule, which is chained to another redirect rule (continuous redirects or too many redirects error).
To find such error, we have to scan all your redirect rules and analyze if there are any conflicts in settings.
If our algorithms find some - Llama Assistant will create a report for you and send using built in chat.
If you enabled
Auto Apply function, then Llama Assistant will have permissions to make changes itself.
Next option is to find a suggestions, based on your current settings using the static functions.
As an example of SEO redirects plugin, it could be a 404 (Not Found) log record, which should be transferred to a new redirect rule.
If there is some, you'll receive a message including report in the built in chat box.
If you enabled
Auto Apply function, Llama Assistant will create a new redirect rule, based on your 404 (Not Found) record.
Preparation for Machine learning
To find any suggestions based on your settings, we have to prepare the data to a machine readable format, before transferring it to a Neural Network.
Basically, Neural Network accepts digits in range [-1 ... +1] only. So all your settings must be classified to "ones" and "zeros". After transformation completed, we can pass this data to our pre - trained Neural Networks.
Machine Learning Predictions
Most of Machine learning models are used as classification or prediction algorithms.
According to the LlamasApps plugin you use - there are different classification groups (clusters). Which include errors, split tests, setting hypotheses and other specific solutions.