The core motivation of Optiwisdom is to develop modular and scalable ML engines capable to work with any industry problem with AutoML philosophy. Instead of having a specific industry-based application, our engine approach enables us to keep developing our technology road map with more efficient and flexible design functionalities. Optiwisdom’s engine products can work together or work in the stand-alone state.
Optiscorer is our scoring engine, can score almost anything based on the AutoML philosophy. We are using various high-tech trending state-of-the art algorithms like XGBoost or Light GBM tuned for our framework and we are continuously updating and adding new features to our mainframe.
Our very simple GUI starts with a simple file loading interface, for our main purpose is to embed it into any system with a Web API. This feature can score the possibility of an employee leaving the company, of a credit card customer renewing his/her card, of various risk scores for a customer to lead to and purchase a product, or the score of an advertisement quality.
Optisegment is the segmentation tool, which can generate both clusters or sets-of entities, such as new customer segments, and/or new marketing segments that human data scientists cannot realize. Customers are using it to generate various types of segmentation such as product segmentation or risk segmentation, in order to find risk groups. This engine is powerful enough to detect anomalies that can identify the outliers after running the segmentation. In addition, it uses trending algorithms like “consensus clustering” to provide better results.
Optimatcher is the matching tool can generate connections between customers & advertisements, customers & products, or employee & jobs and many others. It uses three main approaches namely recommender, matching and association rule mining algorithms together in a very special method that is currently being patented.
Different versions have different features. The engines have four versions. Moreover, the business model includes not only licensing the latest versions but also all versions for covering low budget customer’s project. First versions are the basic engines for entry-level customers. The second version features the “time series analysis” concept that, for example, can yield different scores for a new or an old customer, or for summer and winter changes. In short, these engines can detect various seasonal sales during the year.
The third version adds the scalability of the big data computational concept and scaling features. With the Version Three, we enable the distribution of data and running on parallel structures. But it can also boost up the performance for time-critical jobs as real time computing. Finally, with the current top-level product, Version Four can process streaming data with limited information on data chunks and unknown ranges and classes. Real-time streaming data is critical to real-time operational companies for it allows them to analyze their customers for better business. Optiwisdom the group name of these engines, employs ensemble learning and algorithm tournament for all versions in order to determine the best solution automatically, which %100 AutoML.
Among the automatized steps of preprocessing are feature detection, feature engineering, and auto feature generation. This process, now being patented, represents a totally new approach to feature elimination and dimension reduction at the same time. Offering even more value for the customers, they can combine three engines in any combination like Optiscorer + Optisegment to increase the scoring success or processing time after segmentation. Or they can setup their “customized engines” according to their industrial needs. If the desired outcome is segmentation than Optiscorer can support the Optisegment for even better segmentation. Optiscorer can also support Optimatcher and increase the matching possibilities with scored data. For example, in order to optimize the company’s advertisement network, the matching problem between the customer and the advertisement can be increased by matching the scored customer and the scored advertisement. Adding segmentation of scored customers and segmentation of scored ads to the matching problem of ad networks.
Additionally, the start-up uncovered two new domain application that allows their newly developed engines to collect user experience & performance from the field. The first one is for fraud detection problem. This is Optifraud. It detects fraud by using the three engines together: It uses scoring for the fraud scores, adds segmentation to identify the marginal and outlier data points, and applies matching algorithms to determine the relations between the events generated and the time between them. This product generates the necessary data connection points, as well as the visual demonstration and reporting of the outcomes, and transmits them from the engines to the fraud experts
Most micro-SMEs, SMEs, and larger enterprises are not ready for analytics, machine learning, and artificial intelligence because their data are held in enterprise data silos. In other words, is it fragmented, duplicated, inconsistent, or incomplete – all of which negatively affect data awareness. They have oil but they can’t use it. And they might have other barriers as well, such as
- Lack of the necessary highly educated data scientists to set up, run, and maintain the machine learning (ML) models. One must have data science knowledge to develop the highest-quality model.
- Lack of managing the company’s noisy data.
- Lack of big data infrastructure platform.
- Lack of realizing that data cleaning and enrichment are time-consuming operations.
- Lack of real-time streaming data solutions.
- Lack of an appropriate managerial view to formulate better business decisions.
- Lack of confidence about allowing a third party’s access to their confidential data to devise solutions.
With Optiwisdom Engines, you have your own AI data scientists in-a-box to support your conversion into an AI data-driven company and the latest big data infrastructure platform.
Now is the time to launch your company’s digital transformation with Optiscorer, Optisegment, and Optimatcher. Manage your company’s past data with our ML technology through our self-manner, standalone, and replaceable modular engines. These scalable engines bring new opportunities to ML in off-line, in-memory static, or in streaming data from feature engineering, automating hyperparameter optimization, automated stacking (ensembles), pipeline optimization, and feature engineering to our patented new funnel approach. As a result, you will receive unseen driverless analyses and the best performance predictions for your business.
The content of your data does not interest us, because we do not even see it! Once our engines build the model in a few days, you are done and your confidential data will be stored safely.
In addition, rather than having industry-free AutoML engine solutions, we have end-user products such as AnalytiXR, a next-generation analytics tool for customer relations (CR) and human resources (HR) analytics (CR + HR = XR). This tool provides our patented funnel approach for complex event processing, which you can monitor or even score and predict your customers’ behavior. Its four steps (1) provide your visitors’ scores, (2) understand your customers’ trends, (3) show how your products and/or services meet their needs or requests, and (4) let the AutoML tool AnalytiXR improve their experience. As a result, you will be able to identify visiting customers, leading customers, promoters, and advocators. This tool can both predict the churning customers for each step and output the score tables for further applications, such as creating a promotional campaign or sending SMS or emails.
Optiwisdom has added more features to AutoML to create what we call AutoML+. These added unique and patent-protected special business solutions, are the following:
- Automatically Augmented Dimensionality Reduction: This patent-protected algorithm significantly reduces and transforms your data to an ultra-compact size without any loss and allows you to return it to its original format at any time. This feature gives us computation speed for streaming and also lower memory consumption and requirements.
- Works with streaming data in multiple parallel structures. This feature based to our advanced frameworks and modular engines. We are always using the latest and fastest algorithms and fine-tuning our framework.
- Funnel Display: We use any noisy, dirty, or missing values as input and process self-developed methods, which we then shape it and visualize as a funnel approach. This makes the customers’ operations more meaningful and significant for their business decisions.
- Matching features give your business more power, for our solutions are not limited to scoring; rather, they go one step further by matching the probable sides.
- By using the classification feature in our scoring, which is more than just simple scoring, our customers will obtain more precise scores.