Under normal circumstances, what we call analysis refers to the use of statistical methods of large amounts of data, qualitative and quantitative analysis, interpretation and prediction, and fact-based management to promote the decision-making process and achieve value enhancement.
According to the method and purpose of analysis, data analysis can be divided into descriptive analytics, predictive analytics, and prescriptive analytics.
Descriptive analysis includes data collection, sorting, tabulation, charting, and description of the characteristics of the food being studied. This type of analysis was previously called "reports." Descriptive analysis can be very useful, but it cannot explain why a certain result appears or what may happen in the future.
Predictive analysis can not only describe the relationship between data features and variables (factors that can be assumed to cancel the scope), but also predict the future based on past data. Predictive analysis first determines the correlation between variable values, and then predicts the possibility of another phenomenon based on this known correlation. For example, after seeing an advertisement, a consumer may buy a product. possibility. Although the predictions in predictive analysis are based on the relationship between variables, this does not mean that the causal relationship needs to be clarified in predictive analysis. In fact, accurate predictions are not necessarily based on causality.
Normative analysis is higher-level analysis, such as experimental design and optimization. Just as a doctor will prescribe what actions the patient should take, the experimental design attempts to give reasons for certain things by doing experiments. In order to be able to confidently make inferences in the study of causality, researchers must properly handle one or more independent variables and effectively control other variables. If the performance of the test group in the experimental environment is much better than that of the camera, decision makers should promote this experimental environment immediately.
Optimization is a method used in normative analysis, which refers to an attempt to identify the ideal level of relationship between a particular variable and another variable. For example, we might be interested in identifying the price that is most likely to make the product highly profitable. In the same way, optimizing this method can identify the inventory levels that maximize retail companies to avoid out-of-stock situations.
According to the type of data collection and analysis, data analysis can be divided into qualitative analysis and quantitative analysis.
The purpose of qualitative analysis is to deeply understand the root cause and inducement of a certain phenomenon. Unstructured data is usually collected from a few non-representative cases and subjected to non-statistical analysis.
Quantitative analysis is the initial stage of analysis, and is usually an effective tool for exploratory analysis. Quantitative analysis refers to systematic empirical research on phenomena through statistics, mathematics or calculations. Under normal circumstances, structured data is collected from a large number of typical cases and subjected to statistical analysis.
At the same time, it includes the following types of analysis:
statistics: the discipline of collecting, sorting, analyzing, explaining and presenting data;
prediction: based on existing data, predict the situation of some variables of interest at a specific point in the future;
Data mining: usually use algorithms and statistical techniques to automatically or semi-automatically extract unknown and interesting patterns in large amounts of data;
text mining: the process of obtaining patterns and trends from text in a similar way to data mining;
optimization: while satisfying constraints Under certain circumstances, mathematical methods are used to find the optimal solution according to certain standards;
experimental design: randomly assign subjects to each group. Then use the test group and the control group to deduce the causality that exists in a particular result.
Although a series of commonly used analysis methods are given here, there will inevitably be considerable overlap in the course of use. For example, regression analysis is the most commonly used method in predictive analysis. At the same time, it is also a commonly used method in statistics, forecasting, and data mining. In addition, time series analysis (time seties analysis) is a specific statistical method used to analyze data changes over time, and is often used in statistics and forecasting.