Big data is the basis for personalization

Personalization is the future of business

Modern society is a commercial society. Industrialization has solved the problem of mass-produced goods for us, and thus has promoted the vigorous development of commerce. With the continuous progress of social science and technology, commodities have also started to become more and more diversified and used to meet the different needs of the general public. Take the television as an example, initially only the difference between the size, then you can choose the brand model. Until September 27, 2011, Haier and Lynx launched online custom TV activities; users can choose TV before the TV production size, border, clarity, energy consumption, color, interface and other attributes, and then by the manufacturer Organize production and delivery to the customer's home. Such a personalized service is widely welcomed, and the amount of 10,000 customized televisions in two days was robbed of light. Similar custom services in air conditioners, clothing and other products have been welcomed by users. These examples illustrate the dawn of a business of the future - improving the efficiency of business operations by enabling users to be more satisfied with products that meet their individual needs, thereby shortening the design, manufacturing, shipping and sales cycles.

Big data is the basis for personalization

To achieve a personalized business model, adequate data is the foundation. Bikini manufacturers know that their products have a market on the beach or in the coastal city. Who can expect to Xinjiang and Inner Mongolia, the most favorite men to buy a bikini for their women? This "unspoken rules" hidden in the data, you need to dig to see the sun, as the "beer and diaper" as the classic story . And "big data" goes a step further than traditional data mining. Large amount of data, data types, potential correlation between the data mining the premise of big data. The entire Internet user and all the goods themselves is a large enough data space, coupled with the potential relevance of space, time, weather and so on, the amount of data needed to know each user's preferences is enormous. The more data, the more accurate the user's understanding.

Internet big data processing technology challenges

The big challenge with dealing with big data on the Internet is the ability to handle big data. In order to catch up with the speed of consumption data to generate data, having enough computing resources is a necessary condition. On this basis, the linear expansion of the computational framework, efficient and stable programming and accurate algorithms are the core capabilities of big data processing.

The second challenge is timeliness. Users on the Internet continue to imply its intention of the operation, and only aware of these intentions in time to the user before the next operation to make a valid response, the ultimate user convenience. This timeliness requires that the system's computational framework be run as a data stream. Eventually lead to the system in real-time streaming load, real-time fault tolerance and other issues with the traditional mass data processing technology completely different.

In order to meet the individual needs to a greater extent, but also must have strong enough customization capabilities. On the one hand, although the customization needs of individual users may be small, the number of users is huge and their customization needs are very different. This problem can not be completely solved without the hard work of a few engineers. Need to have as much freedom as the database SQL language to the user, so that the needs of the smaller can be satisfied by a simple operation. Such customization capabilities in the data storage, computing, query, display and other aspects have embodied.

Ali cloud solution - cloud recommended

Whether it is collecting big data computing and storage capabilities, or real-time computing and algorithmic techniques needed to handle personalization issues, it is not easy for webmasters and developers to quickly get around them. Aliyun cloud is trying to reduce the barriers to personalized services, so that more webmasters and developers to enjoy their own low-cost personalized service. One cloud recommendation (http://tui.cnzz.com/) is a typical.

If a website introduces gourmet recipes, users can browse the "Cha Ji mushroom soup", if you can recommend some related recipes, then you can allow users to stay on the site more time to visit more content. In fact, there are many recommended algorithms to find what users are interested in:

· From the user access log which may find that users access to this recipe after 50% of users will go see "Blood Qi Yiwu chicken soup", this phenomenon must have the reason behind, may be a good recommendation.

· Now that users are looking at "chicken soup" recipes, they can recommend other popular "chicken soup" recipes, such as "Mushroom Chicken Soup".

· By analyzing a user's past history of the visit record, may be able to find that the user relative to other users are more inclined to slow fire soup, it should be appropriate to recommend a similar "stewed chicken soup" such a recipe.

· Relative to "chicken soup", "mutton soup" is also a popular soup category, the user may eat chicken soup tired to want to change the taste.

However, to achieve such a recommendation, the traditional approach requires a lot of manual editing. Neither can be done instantly, it is difficult to guarantee results. It's harder to verify that these recommended algorithms produce good enough results on real traffic. A precise recommendation model must make a comprehensive evaluation of the overall effect of the algorithm itself and the preference of the user for the recommended results of various algorithms so as to find a precise recommendation model suitable for each user. Finally allow users to enjoy the recommended booth "Thousands of noodles" personalized service.

To use the cloud recommendation service, simply register the application on the cloud recommendation website (http://tui.cnzz.com/) to obtain a ten-digit application ID such as "1000001234". And the system-generated JavaScript code (as in the above example) can be personalized recommendation results embedded in the web code (as shown below). This process is usually completed in a minute. Subsequent things to the cloud system. It will begin politely crawling and in-depth analysis of the site: identify the page that actually contains the content, and extract the page title, image, etc. as the recommended material. The system will also continue to automatically adjust the model and weight of the recommended algorithm based on the displayed click effect.

Recommended in the cloud management interface, web site developers can customize the recommended location size, the number of recommended content entries, URL range, display and other parameters. The webmaster can also see clicks on recommended booths and adjust the recommended location parameters as recommended to improve performance.

Cloud recommendation service also provides plug-in support for the mainstream website building tools Wordpress, Discuz !, DeDeCMS. After installing the plug-in, the developer can operate and manage various functions recommended by the cloud in the website management interface. For more information, please refer to the website for recommending cloud (http://tui.cnzz.com/).

According to the background statistics, the overall traffic after the website enables cloud recommendation will increase by 10%. This personal service feels like a bank depositors can get the same interest, is the charm of big data show. I believe with the continuous accumulation of data and the cumulative number of users, personalized service in the era of big data can bring more than 10% of the traffic increase such a surprise!