• 목록
• 답변
• 글쓰기
• 게시판 리스트 옵션
  • 수정
  • 삭제

The variety of computer systems, tablets and smartphones is growing rapidly, which entails the possession and use of a number of gadgets to carry out online duties. As people move across gadgets to finish these duties, their identities turns into fragmented. Understanding the utilization and transition between those units is essential to develop efficient applications in a multi-system world. In this paper we present an answer to deal with the cross-device identification of customers primarily based on semi-supervised machine learning strategies to establish which cookies belong to a person utilizing a system. The method proposed on this paper scored third within the ICDM 2015 Drawbridge Cross-Device Connections problem proving its good performance. For these reasons, the data used to grasp their behaviors are fragmented and the identification of users becomes challenging. The objective of cross-system focusing on or tracking is to know if the particular person using laptop X is similar one which uses mobile phone Y and tablet Z. This is an important rising technology problem and a hot subject proper now as a result of this data could possibly be particularly valuable for entrepreneurs, iTagPro device resulting from the potential of serving targeted advertising to customers whatever the device that they're utilizing.

Empirically, advertising and marketing campaigns tailored for a specific person have proved themselves to be a lot more effective than normal methods based mostly on the gadget that's getting used. This requirement is not met in a number of circumstances. These solutions can't be used for all customers or platforms. Without private data concerning the users, cross-system monitoring is an advanced course of that entails the constructing of predictive fashions that must course of many various indicators. On this paper, to deal with this drawback, we make use of relational information about cookies, gadgets, in addition to different data like IP addresses to construct a model able to foretell which cookies belong to a person dealing with a device by using semi-supervised machine learning techniques. The remainder of the paper is organized as follows. In Section 2, we speak concerning the dataset and we briefly describe the issue. Section three presents the algorithm and the coaching procedure. The experimental outcomes are presented in section 4. In section 5, we provide some conclusions and additional work.

Finally, we've got included two appendices, the first one accommodates data about the options used for iTagPro device this task and in the second an in depth description of the database schema provided for itagpro device the challenge. June 1st 2015 to August 24th 2015 and it brought collectively 340 teams. Users are likely to have a number of identifiers throughout completely different domains, together with cellphones, tablets and computing units. Those identifiers can illustrate widespread behaviors, to a greater or lesser extent, because they often belong to the identical consumer. Usually deterministic identifiers like names, phone numbers or ItagPro e-mail addresses are used to group these identifiers. On this challenge the goal was to infer the identifiers belonging to the identical person by learning which cookies belong to a person utilizing a itagpro device. Relational information about customers, units, and cookies was provided, as well as different info on IP addresses and behavior. This score, commonly used in information retrieval, measures the accuracy using the precision p????p and recall r????r.

0.5 the score weighs precision larger than recall. At the preliminary stage, we iterate over the listing of cookies in search of other cookies with the identical handle. Then, for every pair of cookies with the same handle, if one of them doesn’t appear in an IP tackle that the opposite cookie seems, we embrace all of the details about this IP handle in the cookie. It is not doable to create a training set containing each combination of gadgets and cookies as a result of high variety of them. So as to scale back the preliminary complexity of the problem and to create a more manageable dataset, some basic rules have been created to acquire an initial decreased set of eligible cookies for every machine. The foundations are based mostly on the IP addresses that each system and cookie have in widespread and iTagPro device the way frequent they are in different devices and cookies. Table I summarizes the checklist of guidelines created to select the initial candidates.