Excel-based tool enables business analysts to create analytics-powered decision models, “point and click” without coding, using optimization, Monte Carlo simulation, machine learning, and business rules; now includes simple deployment and automatic refreshing of analytic model results to Microsoft Teams.
Frontline Systems is shipping Analytic Solver® V2023 Q1, a new version of its advanced analytics toolset for Excel (Web, Windows and Macintosh), that enables business analysts to easily build models using business rules, machine learning, mathematical optimization and Monte Carlo simulation, and easily deploy those models in cloud-based applications. Among other enhancements, the new release supports easy deployment of optimization and simulation model results to Microsoft Teams, with automatically “refreshed” results when models are re-run.
As an “Excel Solver upgrade”, Analytic Solver can handle virtually any type or size of optimization problem, ranging from a few to millions of inter-related decisions in a single model. And for years, Analytic Solver has offered powerful features for risk analysis using Monte Carlo simulation, and powerful features for training, validating, and deploying predictive models using machine learning. Analytic Solver V2023, released in September 2022, includes an innovative, patent-pending facility for risk analysis of machine learning models – not yet available in other ‘advanced’ DSML platforms.
“We’ve made ‘model deployment’ easy – sharing analytic model results with others in an organization for use in daily operations”, said Daniel Fylstra, Frontline’s President and CEO, “Business users can do this themselves, without needing ‘MLOps’ or expert developers, just using Microsoft Excel and Teams.” The new tools work with Excel on the desktop and in the browser, with deployment to Teams both on the desktop and in the browser, leveraging all the security measures available for Microsoft 365 users.
Read More: Atos Signs Major Contract With Siemens It to Drive Its Digital Transformation Roadmap
Models for Decisions, Not Just Predictions
Unlike data science and machine learning (DSML) tools that are usually limited to building predictive models, Analytic Solver supports the full range of predictive and prescriptive models – the latter are used to support actual decisions, not just predictions. A typical model deployed from Analytic Solver to Teams might specify the least-cost way to fulfill demand for various products or to staff a call center each week, or the best way to blend industrial chemicals based on current costs and market prices. These decisions are carried out by teams, who need to timely receive and use the model results. And Frontline has found via customer surveys that Microsoft Teams is widely used by exactly these people.
Model Deployment and Use without ‘MLOps’
Most companies have a shortage – at best – of staff needed to implement developer-centric ‘DevOps’ or “MLOps’ to support applications that arise constantly within their lines of business. And even if such a solution is offered, the actual needs of those line-of-business applications often mean that the ‘last mile’ is carried out using Excel. Moreover, the key people with ‘domain expertise’ to solve a line-of-business problem are more often ‘Excel literate’ rather than ‘code literate’.
The new facility in Analytic Solver V2023 Q1 supports the end-to-end process of developing, testing, and deploying analytic decision models by Excel-literate business users, without needing developers or “MLOps’. Via simple point-and-click steps, users can send model results from Excel to a Teams channel, and automatically ‘refresh’ those model results – making them available to the key people needing them – whenever the analytic model is re-solved.
When there is a need to move to a developer-centric solution, Frontline’s product line makes this easy: Solver SDK, its toolkit for developers, can be used entirely from code but also loads and solves models created in Analytic Solver for Excel. And RASON, its cloud platform with a REST API for Web developers, can be used entirely on its own but also loads, solves, and even manages and governs models created in Analytic Solver for Excel.
Read More: SalesTechStar Interview with Troy Townsend, CEO and Co-Founder at Zitcha
More Optimization Power Than Ever
There’s more in Analytic Solver V2023 Q1: Its leading mathematical optimization capabilities are better than ever in this release. Linear programming models solve up to twice as fast, and most linear mixed-integer models solve faster – sometimes 10 times faster or more. Nonlinear optimization models can also be solved faster than ever. Significant speed enhancements have been made in the ‘base product’ Analytic Solver Optimization, and in nearly all of the eight plug-in, large-scale ‘Solver Engines’ that work with Analytic Solver, Solver SDK and RASON.
Risk Analysis for Models Created in Other Software
Only Frontline’s software includes patent-pending methods for risk analysis of machine learning models – easy to use with machine learning (ML) models created in Analytic Solver. But many ML models are developed using other software tools. Now in V2023 Q1, Frontline’s risk analysis methods can be applied to ML models created in other software and saved in PMML (Predictive Modeling Markup Language) form – an open standard widely supported by software for machine learning. Users can simply load the PMML model, and some of the data used to train and/or validate the model, into any of Frontline’s tools – Analytic Solver for Excel, Solver SDK or RASON – then perform a risk analysis to assess how the model may perform differently in practice, and what the business consequences will be.