Predictive analytics in Joomla

[Unisoftdev]

Joomla is already picking up a lot of data able to describe the customers' behaviour of visitors, however, what's missing, it could be programmatically used in the most lucrative way for predictive analytics and automized decisions improving conversion rates, design testing, and so on... Machine learning. I think that this is the simplest way how to once again attract the massive interest in Joomla by big companies, thereafter the small followers. At the moment, there's no CMS providing it. When a computer/data scientist works on a similar project, usually it's very time-intensive, there are tons of problems with cross-compatibility and any big companies pay really a lot because the cross-integration with 3rd party software isn't in any way straightforward, it has to cost ridiculous money because there is many programming, setups, and installations. However, that would be a different story in the case of the core of the Joomla CMS, it needs just small changes and portability to 3rd party libraries.

Let's imagine that Joomla can:

• Recognize visitors by their behaviour and offer them different content according to their interests. Each one type could get something else in feeds, etc.
• Recognize customers by browser language settings, etc. It might detect culture and so on... then serve the content according to your settings.
• Recognize which pages aren't interesting for anyone, then prioritize to show in blog posts/menu/feeds something more successful.
• Headers, colours, and design testing in general. For example, you might choose different 5 colours and headers, then Joomla might try all of them for a period of time, count (make statistics) how it works, and then make a statistically based decision which one is the best for the conversion rate and use it forever. The title and also H1 for this topic include: "Predictive analytics in Joomla". Let's say that I wanna try 5 different titles and select that one which drives as most people from Google search results or the category menu, the biggest amount of clicks. However, do we have to test e.g. 100 articles, 5 times per each and make statistics on our own? I think that none can. But Joomla could be doing it and it's not so hard to make it done.
• Etc...

Best regards

[gws]

But Joomla could be doing it and it's not so hard to make it done.
Best regards

Perhaps you would like to contribute to joomla and do it?

[Unisoftdev]

Perhaps you would like to contribute to Joomla and do it?

In practice, you can take it the easy way and just create a few more rows/columns in a SQL database, count visits/clicks for an item per a season(or something else), and then query it (query the biggest values/numbers of visits), just simple stats (I don't block you to expand the idea and go for a graph, but this will be your idea, not mine). I think that anyone of you is technically capable of that. This isn't the same as the adventures when you'll try to put together the Java software developed in 80' years (used in banks/fintechs till this days) with the most modern Python libraries and CMS platforms running on PHP, but nothing natively working together.

Django framework doesn't provide anything as we talk about, it's not a CMS, it doesn't provide even any admin interface (it's just a framework) or predefined SQL, I had to programme it. I have just finished the same things for my personal project that I proposed to you for the CMS, but I was doing it in Python, not PHP. However, there was the same work with SQL and the Python isn't so far away from the PHP code. I use just 3 versions for testings but I will also change it, I'll write more options (multiple tests) 5+ in general, and furthermore for terms in search results (15+?). The Pythonic version on my website:

Screenshot from 2018-12-15 12-55-54.png

It's not an elegantly designed admin interface (what I didn't mind because it's for a single project) but I hope that you can see the multiple titles. I solved it very simply, just expanded SQL fields and wrote short scripts writing the data, making statistics, and the selections (it just queries data). In this way, the scripts will always test it for a period of time, count the total visits, visitor's conversion rate, bounce rate, business leads, and in the end, will select the most successful option without my attendance, everything will go automatically.

[Unisoftdev]

Test-driven evolution vs web editor's choice... This is what I mean that makes the different results.

[darb]

Joomla4.jpg

Its very interesting and you follow the DMN (Decision Model and Notation) standard from the OMG (Object Management Group) for creating rules & decisions by decision --->predictive analytics (triggers/actions) https://www.omg.org/dmn/

See also this topic here viewtopic.php?f=575&t=950945