Data analysis, a core part of data science, can be categorized into several depending on target data or purpose. The former is a technical categorization which encompasses geospatial, time-series, graph, text, image, etc. And the latter is more suitable for understanding the nature of data analysis.
Although multiple versions of categorization exist, we believe the six categories suggested by Jeff Leek of Johns Hopkins University.
1. Descriptive : Profile the data to understand the data using statistical measures such as average, standard deviation, etc.
2. Exploratory : Explore the data to identify correlation among variables.
3. Inferential : Statistically infer whether correlations found in the sample can be applied to the general population.
4. Predictive : Create models out of generalizable patterns within sample data that can predict the value of target variable taking a series of predictor variables as input.
5. Causal : Search for causality among identified correlations.
6. Mechanistic : Check whether identified causality holds in call cases.

In business setting, data analysis usually covers from descriptive to predictive, sometimes to causal. One caveat is that these are not mutually exclusive but gradually related levels of analysis, meaning that exploratory analysis is not possible without descriptive analysis and so on. Therefore, in order to perform well-known predictive analysis, you need to run descriptive and exploratory analysis first.

Higher level of analysis enables high level of usage. The Y axis of the above lists analytic applications, and the X axis shows the types of analysis. As we move to the right side of X axis, higher level of application becomes available. And
machine learning is the key tool for data analysis. Many believe that machine learning is used for predictive or higher level of analysis but that is not true at all. There are a number of ML algorithms for low level analysis, such as most unsupervised ML algorithms.