Today companies are experiencing rapid increase in the volume and variety of data. These can be used to extract useful knowledge, resulting in better decision-making regarding a production process of economic, financial, social/biological sciences, business, and so on.
With the help of mathematics and statistics, it is possible to convert data into findings or insights that will provide a better understanding of the business process, and better yet, they give you a correct answer to a question of business.
“Data Analytics”, “Data & Analytics” or “Analytics”
Regardless which one is your preferred term to describe your business data analysis, all those tasks aimed at exploring the data, with the intention of finding patterns or useful knowledge, so as to optimize or profitable a business process.
Data analytics attempts to answer questions like:
- What will happen to sales volume if they continue the same economic conditions?
- Under the current market conditions what should be the optimal product price?
Note that the questions described above may not be answered through the analysis results from the traditional database systems. In this regard, other tools should be used to perform advanced statistical models to answer such questions.
Converting raw data into knowledge must be done by carrying out the following 3 main steps:
- Descriptive analytics – The objective is to describe what is happening in a particular situation or scenario in a given period of time. For example: Sales and value indicators taken from the previous quarter.
- Predictive analytics – The objective is to predict what will happen in the future based on the analysis of historical data.
- Prescriptive analytics – Go beyond the descriptive and predictive analytics. This seeks to give recommendations or courses of action showing the probability of occurrence of each decision and its possible consequences.
What statistical techniques are used to convert data into useful knowledge for decision making process?
Various techniques can be used depending on the purpose of analysis or business question. Here are some techniques are described existing statistics and a use case for each of them:
- Cluster analysis – Group in one cluster to all those observations with similar characteristics. In other words, two observations in different groups in some sense are not similar. Example: Customer segmentation where each client corresponds to an observation.
- Decision Trees – It is a tool based on decision rules, which uses a logic diagram (decision tree) for performing a classification process. Example: Classifying credit card transactions as “fraudulent” or “not fraudulent” from characteristic attributes of transactions.
- Linear Regression – Statistical methods to identify the relationship between a dependent variable and one or more independent variables Process. Example: To assess the influence of weather on sales of a product and/or service variables.
- Time Series – Statistical model for decomposition and forecasting future values of a series of chronologically ordered data. Example: The forecast volume of customers in a service business cable.
- Operations Research – Enables achieve optimal or near-optimal solution of a complex optimization problem. Example: To evaluate whether the optimal amount of a company stores consumer goods. This analysis can be performed taking into account the turnover of each store, volume and per capita income of the population around, among other social indicators.
- Artificial Neural Networks – They are statistical models that emulate the brain learning process through training, validation and decision. They are mainly used to solve problems of classification and prognosis, such as identification of faces from physical attributes and prediction of economic indicators.
- Support Vector Machines (SVM) – They are more powerful than neural networks. For practical purposes, classification problems solved with neural networks can also be solved with SVM.
What software tools are used for creating advanced statistical models?
In the market there are both free and paid solutions available for data manipulation and generation of statistical models as above and many others.
- SAS – Perhaps the most robust tool for conducting exploration data and statistical modeling. As SAS can connect to different servers and can handle Big Data really well, it is used in different domains in the finance financial sector worldwide. The licensing and maintenance costs are high.
- SPSS – Acquired by IBM, simple interface for loading data and enter the parameters for setting statistical models. It supports model generation by introducing code script.
- Matlab – Application software with a robust library functions classified according to different topics such as time series, optimization, image processing, polynomials, regression, and so on.
- R language – R is the leading tool for statistics, data analysis, and machine learning. It is more than a statistical package; it’s a programming language, so you can create your own objects, functions, and packages. Speaking of packages, there are over 2,000 cutting-edge, user-contributed packages available on CRAN.
- Weka – Excellent tool for building models based on machine learning.
- Orange Data Mining – Data mining from a nice graphical interface from data entry to generate statistical models. It uses a set of interconnectable blocks where the output data of a block are the data set input to the next.
Data analytics is not a software application that can answer business questions or solve business problems magically; however, data analytics corresponds to a set of activities ranging from identifying the business need, the data required to solve the need, statistical models and key business knowledgeable staff to assess the relevance of the results through a feedback process that allows mature models created.