RSS
Логотип
Баннер в шапке 1
Баннер в шапке 2
2015/02/17 00:00:00

Predicative (predictive) analytics of Predictive Analytics

Further development of the world market of the business analysis will go on the way of active mastering of advanced (advanced) analytics, including the predicative (predictive) analysis, creation of simulators and variable models.

The directory of BI solutions and projects is available on TAdviser.

Content

What is predicative (predictive) analytics?

The predicative or predictive analytics (Predictive analytics) is first of all a set of methods of statistics, data analysis and games theory which are used for the analysis of the continuous and historical data / events for the forecast of data/events in the future.

The most known method of use of predictive analytics is an application of scoring models for assessment of solvency of the client at issuance of credits in bank. Any scoring model is under construction on historical data and if in the past, any customer group was convicted of untimely clearing of the credits, and you in any characteristics are similar to this group, then most likely will refuse to you issuance of credits.

However it is not all areas where the predicative analytics is applied, it can be applied to product development, to the choice of potential audience, to the choice of the following product which you can offer the client (Next Best Offer) and a great number of others.

In relation to predicative analytics the concept data mining as the predicative analytics uses partially similar methods is related. The central entity of predictive analytics is the task determination of a predictor or several predictors (parameters or entities which influence the predicted event). For example, insurance companies I select such predictors as age, a driving experience at determination of an insurance premium. The set of these predictors forms model of predictive analytics which predicts a certain event in the future with some confidence figure.

Analysts of Gartner believe that further development of the world market of the business analysis will go on the way of active mastering of advanced (advanced) analytics, including the predicative analysis, creation of simulators and variable models. An opportunity to creation of such models in 2013 in Gartner was called the 15th obligatory block of corporate BI platforms.

The analytics of the class advanced uses statistics, the descriptive and predicative data mining tools (investigation of data), simulators and optimization means. An ultimate goal of use of all these tools – decision making, the solution of business challenges and identification of opportunities for drawing up the best forecasts, identifications of processes, patterns and other patterns.

That the predicative analysis was successful, in Forrester recommend to follow the following stages accurately: goal setting, data acquisition from different sources, data preparation, creation of predicative model, model assessment, implementation of model, monitoring of efficiency of model.

Scheme of implementation of tools of the predicative analysis


Forrester Research, 2013

Scopes

In difference from data discovery of means of predicative analytics are addressed to specialists therefore are not applied so widely. According to Gartner for 2012, only 13% of users of BI widely involve means of the predicative analysis. Less than 3% use such methods as mathematical modeling, simulators and optimization.

Experts consider that you should not wait for mass implementations in this area, but the trend will gradually change. The reason for that – emergence of a phenomenon of Big Data which pushes the organizations to search of new means of information processing. Gartner considers that those companies which will apply advanced analytics to Big Data will grow 20% quicker than competitors.

According to Eric Siegel (Eric Siegel), the expert in the predicative analysis stated in his eponymous book "Predictive Analytics"[1], the scope of the predicative analysis in fact is very wide. He gives 10 most widespread examples:

  • Direct marketing: the task consists in increase in number of responses by data integration about clients of different a web and social sources. The companies can define efficiency promo campaigns, separating potential clients on segments, location or channels of delivery.
  • Predicative targeting of advertizing: any advertiser wants to know what message is the most effective. Advertizing can be shown in the best way online, based on similarity of clicks, and clients will only benefit from giving of more relevant content.
  • Detection of fraudulent schemes: means of the predicative analysis allow to minimize use by swindlers of false schemes of insurance, receiving the credit and so forth.
  • Management of investment risks: means of the predicative analysis allow to estimate the potential of this or that startup or other asset. The method can be used by the companies and for the choice of the partner, the candidate for purchase or even vendor.
  • Customer retention: the predicative analysis allows to calculate a customer behavior and also to consider the negative factors influencing their solutions.
  • Recommendatory services: users can recommend goods or content on the basis of data on the previous viewings, interests or the analysis of comments on Twitter.

Education: means of predicative analytics can be used for providing more effective techniques of teaching.

  • Political campaigns: vote process modeling.
  • The systems of decision-making in medicine: the predicative analysis can reveal on the basis of a set of factors tendency to patients to diseases of type of diabetes, asthma and other diseases connected with a way of life.
  • Insurance and mortgage lending: exact determination of reasonable cover amount in each insured event.

Trade

Forecasting of consumer demand and planning aktsiyiz [2]

  • Forecasting of daily consumer demand at the level shop/goods for 28 days
  • Forecasting of promotional demand
  • Accounting of commodity substitution (cannibalization) during the actions
  • Accounting of changes in the price, seasonal demand izmeneiya
  • Accounting of temperature and weather conditions on the cities of location of shops, sizes of shops and so forth.
  • Forecasting for new goods items, new shops
  • Accounting of opening of competitors

Removal of significant goods items for buyers (Key Value Item Analysis)

  • Selection of goods of the buyers having disproportionately great influence on perception
  • Having revealed these goods items the retailer can influence consumer perception having adapted the price strategy
  • Using aggressive price strategy, retailers can influence traffic, general idea about the prices, profitability, a market share and so forth.

Optimization of the regular and promotional price

  • Recommendations about the optimal price
  • Accounting of restrictions on the calculation, margin, turnover, deliveries and so forth.
  • Price elasticity calculation
  • Recommendations about the time prices for acceleration of sales
  • Direct offers for clients
  • Multichannel sales

Segmentation of buyers

  • Behavioural and marketing segmentation,
  • Purposeful marketing campaigns
  • Analysis of a consumer basket,
  • Recommendations about goods
  • Cross-sales and increase in sales level,
  • Strategy of the best alternative,
  • Prevention of outflow of clients by calculation of a consumer risk

Selection of groups of buyers with similar behavioural characteristics by multidimensional data analysis
Customer Segmentation, Behavioral Targeting, Churn Prevention

  • Increase in conversion on stocks by formation of target groups (segments) of buyers for the directed actions
  • Increases in profitability by recommendations about the level of discounts for different target groups of buyers
  • Increase in loyalty by early identification of buyers with the largest probability of leaving and the subsequent actions (actions)

Predictive analytics on production

  • Analysis and forecasting of impact of influences of factors for products parameters
  • Failure prediction of the equipment - transition from service according to regulations to service on a status
  • Production forecasting of products and energy consumption and resources
  • Online anticipatory notifications about future non-staff situations


For industrial enterprises where processing and understanding of a huge number of data is required and there are high risks at decision making, the predictive analytics is of particular importance[3].

Data on course of technology process are not always used effectively while they can be used for optimization of operational processes and increase in technical and economic indicators of production. Optimization can be executed on any type of production with the serious level of automation, organized collecting and long information storage. For this purpose intelligent systems which can analyze a status of technology process in real time are successfully applied, predict further course of process, determine the level of optimality and, if necessary, change managing parameters or make recommendations to the manager. For the solution of these tasks using means of machine learning the predictive mathematical model of technology process is created. She analyzes input parameters, in real time issues the forecast of course of process and the offer on its optimization. This model integrates with the PCS, MES and ERP systems of the enterprise.

One more task for predictive algorithms is a maintenance and repair of the equipment. Generally the enterprises use the basic mechanisms of control provided by equipment manufacturers. But the potential of these means is limited as they do not allow to analyze the additional factors influencing a condition of equipment and in advance to predict critical situation. Thus, the staff of maintenance department obtains a set of data, but does not know how these data are connected among themselves. As a result reaction from repair services follows only after a hardware failure that leads idle times, and, therefore, additional expenses. The predictive analytics carries out by means of machine learning and artificial intelligence the continuous analysis of Big Data, executes data visualization about a condition of equipment at the moment and predicts scenarios of emergence of hardware failures. Unplanned idle times are as a result reduced, works on TORO are optimized, maintenance time decreases, and the supervision manpower receives the profound analysis of causes of failures of the equipment.

World market

The forecast of Transparency Market Research of 2017 in the 2019th

The market of predicative analytics will reach $6.5 billion by 2019, according to the forecast[4] of November, 2013. According to analysts of this company, growth of the market is controlled such drivers as increase in demand at user analysts and intellectual software for information security and protection against a fraud. Separate rapid development a segment of cloud solutions for the predicative analysis[5] is noted[5].

For comparison, at the end of 2012, according to the same firm, the world market of systems for the predicative analysis made of $2.08 billion, and its annual average gain during the period from 2013 to 2019 will make 17.8%.

The predicative analytics in the industries working with end consumers such as bank and financial services, insurance, public sector, pharmaceutics, telecom and IT, retail is most demanded. 71.8% of volume of implementations in 2012 were the share of these segments. Throughout a forecast period the maximum dose of projects is necessary on the banking sector, financial services, insurance. However, most quickly the number of projects will grow in retail and on production.

Analysts note that growth of cases of fraud, non-payments, mismatch threats to numerous rules and regulations force business to address even more often the predicative analysis for the purpose of creation of the futuristic models allowing to take preventive measures in relation to unfavorable events.

Such different types of the software as the systems of the user analytics, analytics of information security and to campaign managements made about 50% of the market of predicative analytics in 2012. These solutions are used for optimization of organizational processes on sales and marketing, customer management and sales channels, financial and risk of management and so on.

Among the regional markets North America will be the largest market of the systems of the predicative analysis, and here demand for forecast solutions will come from the outside the companies which are actively resolving issues of work with Big Data (big data). For this reason shortly for lease of predicative analytics there will be all key vendors of solutions for big data, including SAS Institute, SAP, Oracle, IBM, Microsoft, Teradata and Tableau Software.

The market at the same time remains is in many respects divided between the largest players: 80% of market size in 2012 were the share of the first five of suppliers. Among other noticeable players Fair Isaac, Tibco, Information Builders, Alteryx, Qlik (QlikTech) and MicroStrategy are noted.

Systems of predictive analytics

Forrester 2017
Forrester 2013

Open source of a system of the predicative analysis:

  • KNIME
  • Orange
  • Python
  • R
  • RapidMiner
  • Weka

Commercial systems of the predicative analysis:

Leaders in the market of the predicative analysis in the field of Big Data

According to[6] for 2013, leaders in the market of solutions of the predicative analysis in the field of Big Data are SAS, SAP and IBM. Examination is also quite strong Tibco Oracle, StatSoft and KXEN whereas perspective vendors are Angoss, Revolution Analytics and Salford Systems.

See Also

Business Intelligence, BI (world market)

Trends of development of the world market of BI

Business Intelligence (Russian market)

CPM (world market)

Big Data world market

Big Data - Directory of systems and projects

Self-Service BI

Data visualization

Cloud/SaaS BI

Open Source BI

Notes