The solution of personalized Internet service in the age of large data

Personalization is the future of business

Modern society is a commercial society, industrialization solves the problem of mass production and promotes the flourishing of commerce. With the continuous development of society, commodities are becoming more and more diversified, in order to meet the different needs of the public. Take the television as an example, initially only the size difference, later can choose the brand model. Until September 27, 11, Haier and the cat in the online launch of custom TV activities, users can be in the television production before the television size, frame, sharpness, energy consumption, color, interface and other attributes, and then the manufacturers organize production and delivery to the customer's home. Such personalized services are widely welcomed, and 10,000 sets of television sets have been robbed within 2 days. Similar custom services are popular with customers in air conditioners, clothing, etc. These examples have demonstrated the future of business--by satisfying personalized needs and using households to get more satisfying products, thus shortening the cycle of design, production, transportation, and sales to improve the efficiency of business operations.

Large data is the basis for personalization

To achieve a personalized business model, sufficient data is the basis. Bikini manufacturers know that their products have a market in the beach or coastal cities. Can anyone expect that men in Xinjiang and Inner Mongolia love to buy their own women a bikini? Such "unspoken rules" are hidden in the data and need to be dug in to see the light, just like the classic story of "beer and diapers". and "Big data" is further than traditional data mining. Large amount of data, many kinds of data and potential correlation between data are the prerequisites of mining large data. The entire Internet users and all the goods themselves is a large enough data space, plus space, time, weather and other potential related factors, want to know each user's preferences, the amount of data required is huge. The more data the more accurate the user's understanding.

The technical challenge of internet large data processing

Dealing with big data on the internet is challenging, and the first is the ability to deal with big data. It is necessary to have sufficient computational resources to keep the speed of consumption data up to the speed of generating data. On this basis, the linear expansion of the computational framework, efficient and stable programming and accurate algorithm are the core capabilities of large data processing.

The second challenge is timeliness. The user's operation on the internet continues to imply its intention, only in time to perceive these intentions, in order to the user before the next operation to make an effective response, and ultimately to bring convenience to users. Such timeliness requires that the system's computational framework be able to operate in the form of data flow. In the end, the system adopts the technical scheme which is different from the traditional batch data processing in the aspects of real-time shunt load and real time fault tolerance.

In order to meet personalized requirements to a greater extent, you must have a strong enough customization capability. On the one hand, although the customization requirements of individual users may be small, but the number of users is huge, customization requirements are very different, not a few engineers will be able to completely solve the problem. There is a need to give users as much freedom as the database SQL language, so that smaller requirements can be met by simple operations. This ability to customize the data in the storage, operation, query, display and other aspects are reflected.

Aliyun Solution--Cloud recommendation

Whether it is to collect large data computing and storage capabilities, or to deal with personalized problems of real-time computing and algorithm technology; for webmaster and developers are not easy to quickly solve the problem. Aliyun is trying to reduce the threshold of personalized service through cloud services, so that more webmasters and developers can enjoy their personalized services at low cost. The Cloud recommendation (http://tui.cnzz.com/) is a typical.

If a website is introduced to the food menu, users browse the "Tea Tree mushroom Chicken soup", if you can recommend some related recipes, then you can allow users to stay more time in the site, access to more content. In fact, there are a number of recommended algorithms to find out what users are interested in:

L from the user access log may be found in the user access to the menu after 50% users will go to see "Blood Benefits Qi Black Chicken Soup", this phenomenon must have its reasons behind, may become a good recommendation.

Since the user is watching the "Chicken Soup" category of recipes, it can be the site of other popular "Chicken soup" recipes recommended, such as "mushroom chicken soup."

• By analyzing a user's past history of visiting records, it may be found that the user is more inclined to other users than the slow simmer soup, it should be appropriate to recommend a similar "stewed chicken soup" recipes.

L relative to the "Chicken soup", "mutton soup" is also a popular type of soup, the user may eat chicken soup is tired of wanting to change the taste.

However, to achieve such a recommendation, the traditional approach requires a large number of manual editing work. Can not be instantaneous, it is difficult to guarantee the effect. Manual editing is more difficult to verify whether these recommended algorithms can produce good enough results on real traffic. A precise recommendation model, the overall effect of the algorithm itself and the user's preference for various algorithm recommendation results should be evaluated comprehensively, so as to find the accurate recommendation model for each user. Finally let users enjoy the recommended booth "thousand people thousand Noodles" personalized service.

Using the Cloud Referral service, only site developers can register for a 10-bit application ID, such as "1000001234", on the Cloud referral site (http://tui.cnzz.com/). Personalized recommendation results can be obtained by embedding the system-generated JavaScript code (for example, above) in the Web page code (pictured below). This process is usually completed in a minute. The next thing to the cloud system. It will begin the polite crawl and in-depth analysis of the site: identify the real content of the page, and extract the page title, pictures, etc. as the recommended material. The system also continues to automatically adjust the model and weight of the recommended algorithm based on the displayed click Effect.

In the cloud recommended management interface, site developers can customize the recommended location size, the number of recommended items, URL range, presentation, and other parameters. Site owners can also see the recommended booth click, and appropriate adjustments to the recommended position parameters to improve the effect.

Cloud recommendation Service also for the mainstream site tools WordPress, discuz!, Dedecms and other plug-ins to provide support. After the plugin is installed, the developer can operate and manage the various functions recommended by the cloud at the Site Tools management interface, please refer to the Cloud Recommendation website (http://tui.cnzz.com/).

According to the background statistics, the Web site to enable the overall flow of cloud recommendation will be increased by 10%. Such personalized service makes people feel like money deposit Bank can get interest, it is the display of large data charm. believe that with the continuous accumulation of data and the cumulative number of users, personalized service in the large data era can bring more than 10% traffic to upgrade such a surprise!