Suggestions from a successful data mining person for data mining graduate students

I used to make some detours on Data Mining Research. In fact, from the origins of data mining, we can find that it is not a brand new science, but a combination of research achievements in statistical analysis, machine learning, artificial intelligence, and databases, in addition, unlike expert systems and knowledge management, Data Mining focuses more on the Application Layer .

Therefore, data mining integrates a considerable amount of content, and it takes a long time to fully understand all the details. Therefore, we recommend that you use About three months to learn about several common technologies of Data Mining: classification, clustering, prediction, association analysis, and isolated point analysis And so on. This understanding is rough, and the goal is to understand what these technologies are used. Algorithm What is it, and under what circumstances should we choose.

After preliminary understanding, it is necessary to enter the selection stage, select a specific direction that you are interested in, and then read the classical papers in this direction ( Overview, main development directions, Application Achievements ). The selection phase may take a long time, for example, one year. At this time, Gradually clarify the breakthrough point, that is, the innovation point of your thesis in the future. . Innovation is very important for research. On the one hand, this innovation is indeed better than the original method, and on the other hand, this innovation is indeed of practical value.

Then, we need to implement our own ideas. Generally, for master thesis, a prototype system is required for testing and the experiment results are used to support the topic of the thesis. The prototype system is the implementation of its own innovation, and needs to be well designed and developed. It should be noted that the establishment of a prototype system is different from the development of a commercial system, and a good theoretical basis must be embodied. That is to say, the prototype system is not simply used to implement functions, but to put your entire set of theories into practice. This theoretical basis will also be included in your paper to reflect the theoretical height of the paper.

It usually takes at least a year to build a prototype system and produce convincing results. Therefore, we need to focus on the core part (the part that reflects the innovation of the paper), the peripheral interface, and so on. We should not put too much effort into it to avoid losing control of the progress.

Finally, the paper is organized and written. We recommend that you gradually write short papers (published in journals and conferences) in the previous phase, such as reviews, system frameworks, algorithm kernels, and applications. In this way, when I finally write my graduation thesis, I have enough content to write better and faster.

The above is a general discussion. In fact, I think the key point is the selection of questions. The quality of the selection depends on your understanding of the current situation of Data Mining Research, your interests and expertise, and the significance of this direction in application. We recommend that you communicate with your mentor and colleagues to make your direction clearer.

As for the employment in the field of data mining, the prospects should be good. If you Interested in research , Such as Microsoft Research Institute, Google, University Research Institute They are good. Practical applications Many large companies include IBM, Accenture, Asiainfo And so on. Of course, some organizations of Party A, such Securities, insurance, and finance Enterprises also need to analyze talents.