Summary and sentiment of data analysis

Niche research two this year, engaged in software data analysis and excavation less than two years. Two years, the niche busy, never summed up their work, today suspended busy footsteps, random writing a few lines of text, right when a sneak summary of the two years of data analysis and research experience and understanding. We encourage each other!

Analysis of the data in fact, it is difficult to say that simple is simple. The difficulty of the analysis is that the initial analysis of a professional field of data is "impossible", the volume of data, the number of kinds, the speed of the update will really make all analysts can not touch the brain. At this time, the most important thing is to look at the data, manual one of the interpretation of data, the beginning of the most basic or to solve what the problem, all aspects of the data, the different attributes of what meaning, the relationship between attributes and so is to interpret the data need to be completed at the beginning of the goal. Of course, this process is painful, everything is difficult at the beginning!

The second difficulty is to find the objective of the analysis, that is, tentative analysis to solve the problem, this is not our niche rookie can solve the problem. At this point, what is really needed is to discuss the study with a professional in the field, or why you need an apprentice at first! (apprentices do not mean finding a master to help you solve problems, but to master the "pattern" of open horizons to help us find the problem to be solved. You don't sound very indebted to me, Marge! OK, the problem is determined, the objective of the analysis is determined, but at this time the abstraction of the problem is not the average person can imagine, if you can understand the problem you are also "Daniel"! This is one of the difficulties in analyzing data. Do not understand the main problem is not deep understanding of the data, or to continue to look at the data manually. But the niche to congratulate you is you upgrade, at this time to see the data is not to solve what, but should be in the heart with the problem, with the analysis of the goal to verify the data, to see whether the data really exist these problems, or to find a solution to the problem angle.

To this point, the following is to solve how the problem, you may need to find the internal rules of the data, it may be necessary to do some statistical verification laws of universality. But to this step should be familiar with the data in this field, analysis is not so uncomfortable (to achieve their own ideas is very interesting), combined with the analysis of the target may be able to propose some technical methods to achieve some technical routes. This is followed by the feasibility and effectiveness of modeling validation techniques. The model may have some parameters to adjust, the values of these parameters will affect the results of the model, so the training data set is very important at this time, but the niche is not a lot of sentiment, after all, the data analysis only for two years, before the training data also just do a few groups, mainly divided into training sets and experimental set. The training set is to adjust the model parameters, and the experimental set is mainly used to verify the validity of the model.

　　

Thank you all to live lonely, can live trouble, read this line of boring words. Perhaps you will criticize two sentences, suggest a few points! Humble learning, high-profile work, sincere people, has always been the principle of dealing with the niche. Thank you for your attention, everyone's suggestions!

Summary and sentiment of data analysis