Predictive analytics is a class of data analysis methods that allow predicting the future of events and objects by analyzing current and historical information to make the right decisions.
Predictive analytics develops together with the data science and it is one of the most promising and rapidly developing areas in IT.
It uses methods of data mining and game theory along with classical statistical methods. Data forecasting is made using the most relevant technologies: machine learning, regression analysis, decision trees. Neural networks are also widely used in predictive analytics while analysing data models
We can accurately assess risks and opportunities and make optimal decisions based on the current situation and the unification of mathematics, statistics, econometrics with business analysis.
The forecast analytics has been widely used in business and digital economics for business processes transformation starting from the automation of data preparation to the use of prediction estimate to make strategic business decisions.
In 2013 SAP, one of the world leaders in the market for analyzing large data, introduced a separate desktop product containing predictive analytics methods called SAP Predictive Analytics.
Picture 1. Preview project in SAP Predictive Analytics
SAP Predictive Analytics
General information
SAP Predictive Analytics is a solution that allows you to collect data and build predictive models on these data that predict the course of subsequent events. The construction of predictive models is carried out by searching for patterns and relationships in historical and current data.
Forecast models can be built using such tools as Automated Analytics (allows you to create simple models, such as classification, regression, clustering, time series models) and Expert Analytics (allows you to use complex analysis algorithms using the statistical analysis language and open source code in the R language )
The special feature of this business solution is the ability to connect SAP Predictive Analytics with productive databases, where operational data is stored. And this makes it easy to create actual data forecasts. A user-friendly interface makes it possible to work with simple predictive models without the recourse to specialists. Therefore, the decision-making process is significantly accelerated
The European Specialized Commission for the Control of Transport of Animal and Animal Products has solved the problem of fraud detection: previously the control was carried out on the individual suspicion of the inspector, which allowed detecting only a small percentage of illegal shipments. After the introduction of SAP Predictive Analytics, checking 30% of containers led to the detection of 85% of violations. It really works.
Functionality
SAP Predictive Analytics allows you to work with data loaded from productive databases (tables and different views are supported SAP HANA, SQL view on SAP HANA, virtual tables Smart Data Access). It is also possible to work with data loaded from flat files (Microsoft Excel files and text files). All this simplifies the work with data sources and makes the work of SAP Predictive Analytics more universal.
Picture 2. Preparing data in SAP Predictive Analytics
At the stage of the data model preparation, it is possible to change the appearance of data source with the help of convenient tools with an intuitive interface:hide and rename measures and analysts, create new analytical objects (populations with time stamp, entities, KPIs, and analytical records).
For this, there are convenient tools with an intuitive interface. Automated Analytics and Expert Analyst tools are used to build predictive data models. In the Automated Analytics tool, the forecasting and analysis of data is performed using several modules:
- The Data Manager module helps in the preparation of analytical data. It is possible to aggregate the event log to unite information from other sources, perform sequence analysis (aggregate events into a chain of transitions), perform preliminary analysis and estimate of text variables.
- The Modeler module offers a wide range of possibilities to create analytical data models: classification models, regression models, clustering models, time series analyzing. It is possible to identify and understanding the phenomenon by a time series as well as to predict its development in the short and long term and the Modeler provides generating association rules based on master data tables and transaction table as well as loading the previously generated data model.
- Automated Analytics is also a powerful tool for analyzing information from social networks. In the module Social there are opportunities to extract and use implicit structural relational information, which is stored in data sets of different types, including social networks. There is a possibility to analyze geolocations and frequent paths in social networks, and you can also download a ready-made model for analyzing information from social networks.
- The Recommendation module allows you to develop a system of product recommendations for customers based on information received from social networks as well as to download a ready-made recommendations mechanism.
When analyzing the social networks of clients of drugstores in the United States, fraud was revealed by patients and doctors. The bases of already known cases of fraud were drawn up, a scheme of connections of patients, doctors and pharmacies was constructed, networks of communications were built. As a result, the effectiveness of fraud detection has significantly increased.
SKYLARK company, which owns the largest restaurant chain in Japan, used analysis of social networks and orders to increase the level of individual marketing offers to customers. As a result, this company has established a system that allows each client to select the best offers at the optimal time for him.
The Expert Analytics tool is a more sophisticated tool for in-depth analysis and more accurate data prediction. The possibilities of forecasting time series, detection of extraneous values, trend analysis, classification analysis, segment analysis and affinity analysis are available. Special Expert Analytics' flexibility is ensured by the fact that Expert Analytics has a wide range of prediction algorithms using the open source statistical analysis language R and memory data collection functions for efficient processing of large amounts of data.
Picture 3. Building a data prediction model in Expert Analytics
Expert Analytics also has convenient data visualization tools that allow you to present analysis results including in such forms as dot matrix diagrams, parallel coordinates, cluster diagrams and decision trees.
Picture 4. Visualization of data
Advantages of SAP Predictive Analytics
- Support of algorithms in the R language
One of the main advantages of SAP Predictive Analytics is the support of user-defined algorithms to forecast data in the R language, which provides flexibility and broad application capabilities when used in non-standard data forecasting cases. Also, an obvious advantage is the set of ready-made open-source prediction models in the R language, which allows you to modify the already prepared prediction algorithms.
- User-friendly interface
SAP Predictive Analytics has a convenient, intuitive interface for modeling, forecasting and visualizing data, which allows not only to reduce the labor costs for forecasting data, but also to provide the ability to predict the consequences of its decisions directly by decision makers without the need to resort to the help of data forecasters, which can shorten the decision-making time and make the company's work more mobile.
- Extensive data visualization capabilities
SAP Predictive Analytics has extensive data visualization capabilities, similar to SAP Lumira. This allows us to visually demonstrate the results of the forecast and clearly see the picture of the consequences of the decisions taken. Graphs and diagrams based on forecast data can be exported in the form of flat infographics, interactive boards, and can also be integrated into reports.
- Integration with productive databases
SAP Predictive Analytics supports data downloaded from productive databases (tables and various views of SAP HANA, SQL view on SAP HANA, Smart Data Access virtual tables). This makes it possible to build productive forecasts based on the most up-to-date information.
- Models tracking throughout their lifecycle
The possibility to track forecast models in the course of their work and change them allows you to retrain and correct models as data changes. Thus, it is possible to achieve maximum productivity and accuracy of forecast models.
- Possibility to import existing models related to business content
SAP Predictive Analytics allows you to use standard data prediction models related to SAP business content. Thus, ready-made SAP solutions can be used at full capacity, providing the possibility of modification and implementation of ready-made cases also in forecasting data.
- Possibility to work with different volumes of data
SAP Predictive Analytics supports a variety of data volumes. It is possible to operate both data from small CSV-files, and very large data sets in SAP HANA. This allows not only to work with any productive data, but also to develop and test models on a small amount of data.
SAP Predictive Analytics is a universal tool to process, analyze, forecast and visualize data. This product allows you to create and maintain predictive models of any complexity based on data of any size from different sources. A clear interface and extensive visualization of forecasting results in a variety of forms contribute to the rapid and accurate adoption of optimal solutions at all levels of management in the company, which is the key to the successful operation and development of the company.
Do you want to get a free consultation about working with SAP platform? Contact JET BI expert