Foreword
In most non-computer professionals and some computer majors, machine learning and data mining are two high-profile fields. In the object, this is a habitual error understanding of "pay tribute" (here I added a lot of Idiology). In fact, both areas are constantly skilled and in-depth in other fields of finish theory and practices, and different are only in penetrating more mathematical knowledge (mainly statistics), behind In the article, I will try to explain these mathematics knowledge to everyone in a more easily understood. From the basic concepts, this paper analyzes their relationships and different, and does not talk about specific algorithms and mathematical formulas. I hope to help everyone.
Several related examples
First, you can list some of the application examples related to data mining and machine learning in life to help you better understand.
Example 1 (associated issue):
Classmates who often go to the supermarket may find that some of our products we have in advance may be placed in the adjacent area by the supermarket. For example, the bread counter will be placed on butter, and there will be an old man near the noodle counter. Such items will make our shopping process more quickly and easily.
So how do you know which items should be placed on one? Also, how much is the probability of purchasing another item in the case of purchasing a product? This is to be solved using the related algorithms of related data mining.
Example 2 (Classification Problem):
In the noisy square, people come around. Carefully observe their appearance, clothes, words and deeds, etc. We will unconsciously discuss this person is Xinjiang, Northeast or Shanghainese. For example, in the 2015NBA Finals just ended, all kinds of authorities will analyze the historical data of the Cavaliers and the Warriors to win the conclusion of the Cavaliers or Warriors.
In the above first example, this is a typical multi-class problem when the geographical classification is classified. In the second example, all kinds of institutions predict whether the Warriors will defeat the Knights to win, this is a two-point issue, and the results have only two kinds. The second classification problem is abnormal in the industry, for example, in the recommended system, whether a person will buy a commodity, such as earthquake prediction, fire prediction, and more.
Example 3 (Cluster Issues):
"The object is gathered, and the person is divided into a shadow, there is a shadow of clustering problems everywhere. Suppose the bank has a number of customers' historical consumption records. Now because the business expansion requires a number of wealth management products that face different people, how can we accurately recommend different financial products to different people through the phone? This is a clustering problem. Banks generally cluster all users, and users with similar features belong to the same category, and finally recommend different financial products to the corresponding category.
Example 4 (Regression Problem):
Regression issues or is also a predictive problem is also a fairly grounding application in life. As you know, the securities company will use historical data to predict the stock price trend of the next time or a day. Similarly, real estate developers will also prize in real estate in different area floors based on the region.
Both of the above examples are typical representatives of regression problems, which often predicts a real value based on certain historical data to the target of a specified condition.
It is believed that the above is easy to understand, you should initially understand the data mining and which problems are applied to machine learning (four types of issues listed here, of course, there are examples such as abnormal detection, etc.), which is solved In the face of a new problem, the three elements of the new problem. The following explains what is machine learning and data mining (ie, what) and their relationship and different points.
Data mining
Data mining, also translates to data exploration and data mining. It is a step in database knowledge discovery (English: Knowledge-Discovery in Databases, abbreviation: KDD). Data mining typically refers to a process of hiding information from a large number of data through an algorithm. Data mining is usually related to computer science, and through statistics, online analysis, intelligence retrieval, machine learning, expert system (relying on past experience law) and pattern identification, etc.
It can be seen from the above definition that data mining is a more biased application relative to machine learning. In fact, data mining is a wide cross-discipline involving a wide range of crossings, as long as we clear the business logic, you can convert the problem into mining issues.
Data mining process generally includes data pretreatment (ETL, data cleaning, data integration, etc.), data warehouse (can be DBMS, large data warehouse, distributed storage system) and OLAP, using various algorithms (mainly machine learning Algorithm) Mining and final assessment work.
Briefly, data mining is a series of processing, and the final purpose is to dig out information you want or unexpectedly harvested from the data. The following figure shows a wide range of applications of data mining.
Be
Machine learning
In the previous section, we initially discussed the relevant concepts of data mining. In this section, we continue to discuss the basics, learning methods, common algorithms of machine learning.
The problem that the machine learning this discipline is that how the computer program automatically improves performance as experience accumulation. - Tom Mitchell
The above definition is the definition given in Tom Mitchell in its work "Machine Learning". This definition is simple, but it contains too much things.
Popularity, we write a program to let the computer conduct a learning process until you reach a satisfactory level. So what is the purpose of learning? How to learn? How is satisfied and how is it defined?
Usually, suppose, our goal is a Function F, we will provide a certain training data to your computer to learn training, every study will train a Hypothesis H, when H and F continue to learn more and more When h is getting more and more satisfactory. The degree of satisfaction is measured by error E (a different way of different situations). Simply simple, machine learning is the process of finding a suitable target function through data training. At present, the machine learning discipline has applied a large number of statistical knowledge, and we also call them a statistical machine.
Let's explain a few concepts that must be known.
learning method
According to the different data type, it is different from the modeling of a problem. The algorithm is classified according to the learning mode is a good idea, which allows people to consider the best results based on the input data when modeling and algorithm choosing. In the field of machine learning, there are several main learning methods:
Supervised Learning
Under the supervision learning, each group of training data has a clear identification or result, such as "Xinjiang", "Shanghai", "Dongg", etc. in the region, such as the "Xinjiang", "Dongful", etc. belongs to the region. When establishing a predictive model, the supervisory learning has established a learning process, comparing the predictive results with the actual results of "training data", constantly adjust the prediction model until the prediction result of the model reaches an expected accuracy.
Be
The classification problems and regression problems in the above examples belong to the supervision learning category. The commonly used classification algorithms include: Decision Tree (refer to my previous article), Native Bayesian Classifier, based on the support vector machine (SVM) classifier, neural network method ( NEURAL NETWORK, K-nearest neighborhood (KNN), etc.
2. Non-supervised learning (Unsupervised Learning)
In non-supervised learning, the data is not identified, and the learning model is to infer some intrinsic structures of the data. The association problems and cluster issues in the previous four examples belong to the category of non-monitoring learning. The common algorithm in the association problem includes Apriori (this algorithm based on the previous article), FP-GROWTH, and ECLAT, and the most classic algorithm in clustering issues is K-Means.
Be
3. Semi-Supervised Learning (SEMI-Supervised Learning)
In the semi-supervising learning method, the input data section is identified, and the part is not identified, which can be used to predict, but the model first requires the inner structure of the data to make a reasonable organizational data for prediction. The application scenario includes classification and regression, including some extensions of the commonly used supervisory learning algorithm, first attempting to model unomailable data, and predicting the identified data on this basis. Laplacian SVM.) Or the like is a graph inference algorithm (Laplacian SVM.).
Be
4. Strengthening learning (Reinforcement Learning)
In this learning mode, input data is used as feedback to the model. If the input data is just as an inspection model, the input data is directly fed back to the model, the model must be on this. Make adjustments immediately. Common application scenarios include dynamic systems and robot control. Common algorithms include Q-Learning and Temport Learning (Temporal Difference Learning)
Be
Data mining and mechanical learning relationship
On the above, we introduce the basic concepts, applications, related algorithms of machine learning and data mining. Next, continue to discuss the relationship between the two and the difference.
Statistics - 1749
Artificial intelligence - 1940
Machine study - 1946
Data Mining - 1980
From the development of history, it can be seen that data mining is an emerging discipline, which is built on a strong knowledge system, using a large number of machine learning algorithms, and according to the narrative of the previous section, data mining has also used a series of Engineering technology. Machine learning is a discipline that is supported by statistics, which does not require consideration of application engineering technologies such as data warehouses, OLAPs.
Summarize
Machine learning is a more biased theoretical discipline, and its purpose is to let the computer constantly learn to find a hypothesis H. Data mining is an application discipline that uses many knowledge including machine learning algorithms. It is mainly to use a series of processing methods to dig information behind data. Technology area
Ma Huateng proposed "Tencent Chupo Ecological Partner Program" to build a digital ecological community
Can the company now interact with humans by artificial intelligence?
Which elements of artificial intelligence changed the phone?
After splitting Windows, is Microsoft really play the return of the king?
Huichun Technology: Display IOT, the artificial intelligence field is a major breakthrough, strong push gesture recognition!
Our other product:
