Development of recommendation system for e-commerce through social networks.
Traditional business has been transformed by the arrival of E-business. In fact online selling represents now 48% of the sales. This new vector of selling has become more and more challenging for shops.
As the E-business went famous, the means to stand out have to evolve. Recommendation generators have been implemented. These services can analyse every purchase of each customer, detect global trends and create a user profile to describe buying patterns among the users.
The globalization of these technics tends to create better tools to predict user needs in order to keep with the competitors. During our project we intend to find a new way to make recommendations and not only rely on the analysis of the customer history. To achieve this we used a new actor of the internet: social network. It is well known that the users of social networks exchange personal data (that cannot be use directly as commercial uses) and personal feelings on different subjects.
We developed two algorithms to process user’s data. The first one is a traditional history based recommendation algorithm while the second will treat data collected on Facebook.
Using API’s we can analyse comments or likes concerning products that we sell and then define users’ preferences. Results from that process are mixed with the customer history in order to build the final products recommendation list for each user
In order to evaluate our solution we decided to create an E-business website using PrestaShop platform. PrestaShop is an Open Source base to create quickly E-business ready-made shops online. It allows to integrate different modules created by the users of the platform. We included to our recommendation module a virtual model to try out every suggested products.
After the integration on PrestaShop we tested the performances and reliability of our solution.
Our results are based on the analysis of the algorithms when processing ten to thousands of product and ten to thousands orders. We observed that the factors that could increase drastically the processing time is the loading and analysis of the data. However, 43% of European E-business websites belongs to small companies (less than 250 employees). They add about 10 to 100 product a year. That is why we can consider that our processing time is good enough for production. However, we could not take into account the fact that the two factors could not be modified or far from reality: the fact that we did not test multi-user requests and we did not have a server with better capacity than our personal computer.
We are well aware of that limitations, this is also why we chose to use PrestaShop. Its community is really active and the feedbacks which could be done would certainly allow us to correct our future versions.
What we have done is not lost, it could be used again and improved. New orientations could be done with this project. The agility of our models allow any company to exploit the trends we extract. The company would be aware of its own company’s brand reputation.