"Identifying customers' preferences that they themselves were not aware of through recommendation algorithm using machine learning."

 There is a service that always comes up when we talk about successful personalized recommendation features. Netflix, which has more than 100 million members worldwide, puts a great deal of effort in personalized content recommendations. Netflix recommends content that not only users want but also content they didn't know they wanted through recommendation algorithm based on machine learning. For example, if a user repeatedly stops watching a movie at a scary scene that involves ghosts or monsters, the algorithm learns this preference and excludes any content that has similar scenes from the recommendations list. An employee at Netflix said, "Recommendations are made based on the habits of the users. Users will experience well-chosen content recommendations that meet their preferences the more he or she uses our service."

 "If a customer bought a lipstick, wouldn't she need a hand mirror as well?"

 Amazon is known to have an excellent recommendation system along with Netflix. Amazon makes recommendations of products by analyzing the purchasing patterns of its members. It is known that 35% of Amazon sales come from recommended products, making it the number one contributing factor of Amazon's growth. Among Amazon's recommendation systems, A9 makes product recommendations by deducing customer preference based on his or her recent data entry after building an item matrix that defines the correlation among different products. For example, if a customer is searching for a Samsung TV, the recommendation system immediately recommends products that are thought to have correlation with Samsung TV, such as a TV home theater or HDMI cable. Item matrices for millions of products and tens of millions of customers are required for this to work. And such analysis and recommendations are made possible because of the advancements in big data technologies and the subsequent ground-breaking evolution of analytical capabilities.

 "Predicting news with a higher chance of getting clicks through learning and analysis."

 For portal sites that offer 24-hour news service, recommendation is a core technology. That's because the more the site recommends news that will get more clicks by understanding the preferences of users, the more visitors it will get. It is also convenient for users to get appropriate news recommendations that they are interested in. The Real-time User Behavior-based Interactive Content recommender System (RUBICS), which is a personalized news recommendation system of Kakao, was introduced in June 2015 and has evolved into what it is now through machine learning and AI technologies. Kakao explains that this system can precisely predict news that will get more clicks through learning and analyzing by weighting recent news, adjusting weighting of already-clicked news and reflecting user preferences, and moving away from the limitations of recommending news that are similar to the news a user clicked on after login or pushing news based on gender and age.