Making online recommendations: much easier thanks to software

Ferris wheel

Website visit

A website is an important means of getting into contact with new customers. Visitors are looking for information, contact, products or services that the company supplies. When a visitor visits a website, they will look for the desired page, usually via several other pages. Is it possible to predict what the visitor is looking for? Can certain pages be recommended?

Machine learning cycle

Data Mining

Each visitor has certain characteristics, such as sex, age, interests, city of residence, etc. Pages can also be subdivided into categories, such as information pages, contact pages or shopping pages. Making predictions and recommendations requires lots of data from page visits, so much that it is impossible for people to analyse all of it. For this purpose, a technique originating from data mining is used; Machine Learning. Machine Learning makes it possible to find patterns and useful information from a large quantity of available data.

Machine Learning is a tool that runs in the Microsoft cloud environment. This tool makes it possible to make predictions and recommendations by processing data and analysing it by means of a wide range of algorithms.

Yannick Woerdman

Recommender systems

Recommender Systems are systems used to generate personal recommendations for the visitor. These can potentially answer the aforementioned questions. These systems make use of two filtering techniques: collaborative filtering and content-based filtering. The first technique is based on information from page visits and attempts to discover what relationships exist between various visits.

The second technique is based on data from visitors and characteristics of pages by linking these to each other. A combination of these two methods is referred to as a hybrid recommender. This method can be applied in Machine Learning of Azure.

Machine learning cycle

Machine Learning

A recommender model is trained with 3 data sets;

  • a user-item-rating set,
  • a user-feature set
  • and an item-feature set.

The user-item-rating set is a data set with 3 columns, these include visitors, pages and a rating in the third column; the higher the rating the more the page is valued by the visitors.

The user-feature set is a data set with all visitors and their characteristics. The item-feature set is a data set with all pages and their characteristics. When the recommender model is trained with these data sets, it is possible to make predictions and recommendations for each new visitor on the basis of his data and previously visited pages. Data from a new visitor on a website can be exported in real-time to a machine learning experiment. In this experiment, the trained recommender model will make predictions or recommendations for the visitor. These results can also be linked back to the website in real-time via the Machine Learning web service.

With the right data and the right software, it is relatively easy to create a recommender system for your customers or visitors. Machine Learning makes use of drag and drop modules, making it easy to use for everyone.

Yannick Woerdman

Data is becoming increasingly important

Data science is becoming increasingly important. It allows us to uncover facts and numbers that would otherwise remain invisible to a company. It becomes possible to recognise patterns, make valuable predictions and discover deviations in time. Data becomes more important every day, so don't leave it lying around. Do something with it, and Get Smarter Every Day!

We get inspired by curious people

Our goal is for people to Get Smarter Every Day. 

John van Beek

Get Smarter Every Day

How can we help you?
May Iquality store your contact information for future use?

Read more about our privacy statement.

Thank you for your message

We will get back to you as soon as possible.

Oops, something went wrong.

Sorry for the inconvenience. Please try again later.