How companies are using machine learning/ AI to predict your buying behavior from your social media

Have you ever thought why big companies like Google, Facebook provide their services absolutely free of cost? All these companies have such a huge user base that even if they charge minimal price for their services they can generate billions of dollars in revenue everyday. So how do they generate the revenue and what do they use to get that money? The simple answer will be YOU. Tech giants like Google and Facebook use their users as a data point to collect the personal data that they share while using their websites. You must have heard about Facebook being in the news for selling the data of millions of its users to private organisations like Cambridge analytica.

In this article we will see how tech companies use your data available on social media to train their sophisticated ML algorithms to predict your buying behavior.

What is the overall model for prediction?

Services like Google, Facebook, YouTube generate revenue mainly by showing advertisements to its viewers. And to improve the conversion rate of the ads they must predict your buying pattern and show you the ads of only those products which you are interested in buying.

So the companies use ML algorithms, Neural networks etc. to predict your behavior from the data that you share at the website.

For example, if someone searches about a smartphone on Google or Facebook, they will get more ads related to smartphones. You can try this experiment with your own system also.

A detailed analysis of the prediction:

Now let us dive into the more technical things that happened during the whole process:

Step 1: Data collection and processing:

First of all the data is collected from the search engine for the social media websites. A part of the data is used for training the neural network and the rest of it is used to test it.

The next step is feature extraction. In this process the data and the respective users are grouped and labelled into different classes depending on the properties of the data e.g. geographical location of the user, age group, type of the product searched, number of times searched etc. Now the processed data is used to train the neural networks.

Step 2: Training the ML algorithms:

The next step is using the organised data train ML algorithms. Mainly two types of algorithm can be used for this:

1. Long Short Term Memory(LSTM):

It is a kind of recurrent neural network in which the output from the last step can be used as the input for the current step. The main advantage of LSTM is that it can retain the data for a long period of time and can derive a lot of information from a small set of data.

2. Reinforcement Learning:

Using this algorithm the computer can give the most optimized prediction using real time feedback. For example if the system shows more appropriate ads to the viewers the accuracy of the system increases for the next recommendation.

Step 3: Using the predictions to show ads:

Now the algorithm generates some new information depending on the user data which is used by the system to show the products that the user wants to see. Feedback mechanism also works in this case where the click through rate ( ratio of the number of clicks on the ad to the number of times the ad is shown) is used to determine the efficiency of the system and fine-tune the algorithms.

What If I am not on social media?

Researchers from University of Vermont have shown that the algorithms have become so sophisticated that the behavior of a person can be predicted even from the friend circle of that person, even if he is not on social media. In this case the prediction meaning happens using the approximate geographical region and the buying behavior of the friends of that person. So there is only a little chance to escape the web!

Conclusion:

With the growth of behavioral economics the question of data safety has always been a concern to the authorities. Some governments have also worked as companies for their aggressive strategies for using the user data. So, should we be concerned about how our data is being used? The answer is YES. While you cannot escape the Web fully, you can definitely monitor how the companies are using your data. And thanks to the strict guidelines from the government you can rely on the companies to use your data for development purposes only.