PolyAnalyst Visual Scripting Platform for Data and Text Analysis

Polyanalyst is a platform for visual development of data and text analysis scripts that does not require programming skills to work.

The system is an end-to-end processing and analysis tool. It is able to upload information flows from all commonly used heterogeneous sources (standard files, databases, social networks, etc.) and, having carried out their intelligent processing, present the results obtained in the form of customizable user reports.

2022

PolyAnalyst - Russian low-code enterprise data and text mining platform

The PolyAnalyst platform, developed by Russian software manufacturer Megaputer, is an analytical low-code system and intelligent solution development environment. Work with the platform is carried out using a mechanism for graphically building an analytical script consisting of a sequence of functional nodes. Thus, even users who do not have special mathematical knowledge and programming skills can independently create complete complex solutions for analyzing data and texts, as well as optimizing and automating business processes. Read more here.

Integration with Luxms BI

GK Luxms Megaputer and in 2022 announced a successful pilot in depth multivariable text analytics (). NLP

A joint solution based on Luxms BI and PolyAnalyst can be focused primarily on an impressive layer of useful information contained in text data from the Internet, and will allow you to identify and track problems that arise both in individuals and in entire client segments.

2020

Forecast model for the development of the COVID-19 virus infection pandemic

The Russian company-developer of analytical products Megaputer Intelligence published in May 2020 its own forecast model for the development of the COVID-19 virus infection pandemic in Russia. The model is clearly visualized and is an interactive map of the forecast of reaching the peak of new cases of COVID-19, as well as the total number of active, fatal and cases of recovery. The results of the forecast are presented both in the general country and for each individual subject of the Russian Federation.

For all regions, there is its own daily forecast of the number of confirmed and active cases of the disease, the number of recoveries and deaths. As a result of the analysis of pandemic data by machine learning algorithms, a statistical picture of the course and spread of the disease was obtained. For each indicator, the algorithm forms two options for the development of events: realistic and pessimistic. The second is constructed by increasing the distribution variance parameter. This is done in case of a decrease in the level of self-isolation caused by the onset of the May holidays.

To generate a forecast, the system analyzes historical data obtained from open sources. The model uses machine learning and predictive analysis algorithms that take into account the diffusion of infected citizens between regions in conditions of unhindered movement, as well as the severity of quarantine measures in a particular subject. In a situation where the key epidemiological characteristics of COVID-19 are not yet reliably known, the chosen approach to prediction through AI can lead to the most accurate applied results.

The forecast methodology is based on a combination of different techniques. Combining various approaches is necessary since the forecast is built for individual regions, which differ greatly in the degree of development of the epidemic, demographic indicators, and in the degree of reliability of the data. A wide range of techniques are used - both optimization of phenomenological models, and stable but rather rough methods such as decision trees, and accurate algorithms such as convolutional networks. The procedure for combining them is also the subject of optimization - hyper-learning or "learning learning."

The presented forecast provides answers to many applied questions about the fight against the pandemic. For example, what load the health care system may face in the coming days, and where quarantine measures can already be mitigated. Also, by comparing the course of the incidence pattern in the region with the restrictive measures introduced, it is possible to draw conclusions about the effectiveness of certain methods of combating the disease.

The analysis took into account publicly available data obtained from open sources, such as Rospotrebnadzor and the portal http://coronavirus-monitor.ru. The results of the analysis are a forecast based on historical data, which should be taken into account only after a thorough, in-depth study, taking into account the data on the course of the disease obtained from other sources.

Basic Description and Functional Tasks

The comprehensive analysis process is built up of stages of data cleaning and preparation, research using artificial intelligence algorithms and text analysis in 16 languages, as well as publishing results in an interactive web interface.

PolyAnalyst is used as an open development environment. It includes elements of the Low-code platform, which, together with the mechanism for visual design of an analytical script, allows users who do not have special mathematical knowledge and programming skills to independently create complete multi-step solutions for data analysis and automation of workflows.

• Structured data analysis.
• Analysis of text data: classification, clustering, extraction of entities and facts, identification of trends.
• Development of data automation solutions:
• Monitoring the quality of customer service by analyzing messages sent by employees of call centers;
• Legal expertise of contracts;
• Dispatching the flow of incoming documents;
• Processes and procedures of industrial enterprises for the use and analysis of text documents (contracts, accounts, financial reconciliations, invoices, complaints, dispatch logs, reference books, NSI reviews, correspondence of employees, transcripts of negotiations, etc.), in large volumes passing through various functional units: legal, financial, sales, marketing customer service, information field analysis, logistics, business process support, etc.
• Visualize analysis results and design reports.

Expected Implementation Effect

• Refusal of costly manual work on data analysis;
• Automation of work processes involving interaction with information flows and documentation;
• Improving the efficiency of management decisions by increasing their awareness;
• Aggregating information from all of the multiple sources used;
• Improved reliability of financial/commercial settlements and forecasting.