2.3 Tasks
2.3.6 Select Techniques for Analyze Data
Guide to Business Data Analytics
The following is a selection of some commonly used analysis and analytics techniques applicable to the Analyze Data domain. The following list of techniques does not represent a comprehensive set of techniques used by an analyst in the Analyze Data domain but presents a small, but useful, set of techniques that can be used.
| Techniques | Usage Context for Business Data Analytics | BABOK® Guide v 3.0 Reference |
| Business Case | Used to understand the high-level needs of the business and align the analytics effort to qualify the desired outcome. | Chapter 10.7 |
| Decision Analysis | Used to understand the multiple decision threads and the rationale behind following a particular course of action. For example, the decisions taken by data scientists for data sampling, data transformation, choice of model, and evaluation criteria are validated against the business and statistical parameters. | Chapter 10.16 |
| Financial Analysis | Used to support the decision process by understanding the costs, benefits, financial impact, and business value. | Chapter 10.20 |
| Key Performance Indicators (KPIs) | Used to evaluate the relevant metrics, KPIs, and model criteria to establish the most accurate representation of evaluation parameters for the analytics model. For example, while determining the objective/cost function of a predictive analytics model, KPIs and metrics need to be translated to the mathematical model correctly. | Chapter 10.28 |
| Observation | Used to understand and analyze the data activities and processes to uncover any information that may impact the success of the analytics initiative. | Chapter 10.31 |
| Reviews | Used to understand the whole process of data analytics versus simply evaluating the outcome of the analytics initiative. | Chapter 10.37 |
| Risk Analysis and Management | Used to record and control the inherent risks and assumptions originating due to a certain approach taken for the analytics initiative. | Chapter 10.38 |
| Scope Modelling | Used when re-scoping is needed during the initiative when the analytics objectives, the data, choice of models, or evaluation criteria change. | Chapter 10.41 |
| Data Journalism and Storytelling | Used to communicate the actions and the results of the data analytics initiative to stakeholders in the Analyze Data domain. | N/A |
| Descriptive and Inferential Statistics | Used to understand the underlying data patterns and signals during exploratory data analysis and modelling. Descriptive statistics is primarily used to describe the data in a more cohesive manner. Inferential statistics is used for predictive and prescriptive modelling (for example, a bayesian inference model). | N/A |
| Technical Visualizations | Used to understand the underlying patterns and signals from data in a visual format during exploratory data analysis and data modelling. Technical visualizations are used to analyze the data, while business visualizations are used to interpret and report results. |
N/A |
| Machine Learning (ML)/ Deep Learning (DL) | Used to predict or prescribe outcomes. Data scientists understand the mathematical constructs used in ML and DL to achieve a better model performance. Analysts understand and communicate the characteristics of different models. For example, a Naïve Bayes model can be used effectively as spam detection with a cheaper cost of implementation and less data volume. | N/A |
| Optimization | Used to derive the best possible business outcome where a number of constraints exist. Analysts identify constraints and assess whether it is considered during data analysis and modelling, For example, linear programming in a simple production decision to complex gradient methods for weight optimization in deep learning problems. | N/A |
| Simulation | Used to derive and demonstrate possible business outcomes when there is a lack of observed data, a high degree of uncertainty, or an extremely high number of modelling parameters are present. Simulation can be effective where the problems may not be solved adequately given the time, schedule, cost, or computing constraints. Even with deep learning and big data technologies, it is sometimes difficult to accurately determine solutions analytically. In such cases simulation is used to solve a problem heuristically. Prescriptive analytics and specifically reinforcement learning problems heavily utilize simulations, for example, monte carlo simulations can be used to generate “good enough” models for estimating a portfolio risk in investment banking and risk management. |
N/A |
