Version 1.3
Introduction
Othot is pleased to announce the release 1.3 of the Othot platform. This release adds significant new functionality to the platform as well as improved performance, quality, and security. This document details the changes, additions, and unseen improvements to the platform.
If you have any questions regarding the contents of this document or require additional information on any new feature or change to the platform, please contact Othot Customer Support at support@othot.com.
Shared Views
This release adds additional functionality to the creation of views within an HIQ. Users can now provide additional details about the view definition, see the filters that will be included, and most important, share this view with other users or groups.
Othot Customer Support will be providing a detailed document to help customers understand the differences between the previous release and this release in regard to how to save and share views.
WorkQueue
A new module named WorkQueue has been added to the solution. This module provides a new interface and integration channel to build task lists from prediction sets, what-if results, and optimization outcomes that can be shared with team members or systems to execute the recommended actions. Othot realizes that value from predictive and prescriptive analytics only comes when organizations are able to easily consume the data in workflows and decisions to make positive impacts.
Users, groups, or systems can be assigned tasks created by “Insights” users. These tasks can be simplified lists of data that may include student identifiers and the recommended aid allocation that was prescribed using the Othot optimization.
Additional details and training on how to effectively use the WorkQueue module will be provided to all customers post release.
Additional Predicts
The Othot platform and data science engine are now able to provide additional predictions within a single High Impact Question (HIQ). Additional predictions can be used to predict likelihood of other outcomes such as likelihood a student will persist through the term of their degreed education or the likelihood a prospective student will apply for enrollment. Othot Customer Success will work with customers to determine the additional best fit predictions for their HIQs, if any.
Multiple File Merge
Customers can now upload multiple files for train and predict operations and the Othot engine will merge and even aggregate the data to generate a new merged file for predict and train. The merge logic uses powerful Othot Data Science Language (ODSL) merge definitions and the HIQ can be configured to have different file inputs for train and predict operations. Customers can even directly upload their files to the customer data management area and these files can be tagged to be directly consumed by the engine.
Most Impactful Feature
A new column can be included in customer HIQs to show the most impactful feature that can be changed with a What-If or Optimization. When a prediction does not include any What-if features this column is blank. Customers can use this new column to quickly see which groupings of predictions can have their outcomes changed by specific features.
Expected Outcome Aggregations
The platform now includes two additional aggregation options. One is the expected average and the other is the expected sum. When enabled on specific features, the aggregations will show the expected outcomes for these features considering the prediction or the what-if predictions. For example, this new aggregation can be used to view the expected average SAT score of a set of students based on the current predictions for that HIQ.
Likelihood Factor
Customers can now add a new column to their Insights grid to show the likelihood factor. This factor takes into account the base or average prediction for all data from the training set and displays a multiplier derived from this factor for each prediction target. Customers can use this factor to get a sense of the likelihood of conversion of a student compared to historic data. High likelihood factors would suggest a higher likelihood of conversion where a lower likelihood would suggest less likelihood. Factors can be particularly insightful when the average prediction outcomes of a model are very low.
Groups
Customers can now create groups in their administration console. Groups can represent teams or related roles within the organization, such as Recruitment Team. Groups can be used in view sharing and in WorkQueue task assignment. Othot will continue to expand usage of groups within the platform in future releases.
Data Loss Protection Filter
The Data Loss Protection (DLP) feature within an HIQ allows customers to specify data features that can NOT be exported from the system. These features may be Last Names or other personally identifiable information (PII) that customers wish to control. You can contact Othot Customer Support to determine which features in your HIQs you would want to include in the DLP list.
Life Cycle Correction Factor
Othot’s very powerful Life Cycle modeling can provide much more accurate predictions within a single Life Cycle for a prediction set. However, Life Cycle modeling introduces factors that can skew overall aggregate predictions. The Othot Data Science team has developed a proprietary correction algorithm that can be turned on within a Life Cycle to correct for these aggregate effects. Othot’s Customer Success team will work with customers post release to determine the best use of this factor.
Terminal Node Life Cycle Assignment
Life Cycles can now include special logic for final or end states in Life Cycles. These end states can represent known outcomes such as Application Rejected, which would result in a zero percent chance of student enrollment. Terminal Node Life Cycle assignment enables improved overall aggregate predictions.
Regression HIQ Support
This release of the Othot platform and engine now supports regression based HIQs. This new capability enables HIQ predictions to be numeric outcome based. For example, HIQs may be developed to predict the amount of discount a customer may require to buy a product or the amount of time it will take to fill a position within an organization. Othot is excited to work with customers to develop additional HIQs that will provide value using this new regression capability.
Platform Scale
This release substantially improves the platform scale. Previous releases required a delay to render HIQ results when prediction sets increased beyond 20,000. With architectural changes, the platform can now scale to these larger HIQs with limited delay in result presentation. This release also optimizes limited bandwidth connections to the platform so customers with slower connections will have far shorter wait times as they interact with the platform.
Improved and Expanded Integration
The 1.3 release of the platform includes an updated and expanded release of the Othot Scheduler as well as expanded APIs to interact with the platform. Customers can use this expanded functionality to integrate from and into multiple other systems. This release also increases the scale capabilities of the scheduler to consume very large files for training and prediction operations from SFTP locations.
Other Enhancements
As with every release many existing features have been improved or expanded. A few of the more notable items are:
- Improved Reference Data Visualizations
- Expanded Engine Logging for Othot Data Science/Support Troubleshooting
- Improved Optimization Interface and Flow
- Additional Configuration Capabilities for Likelihood Visualizations
- Improved Date Sorting and Filtering
Data Science Engine
Driving the Othot solution is the powerful Othot Data Science engine. In release 1.3, Othot is launching the next version of the engine in the platform. One of the most notable additions in this release is the expansion of the Othot Data Science Report. This report provides our Data Scientists with extensive data and details related to the HIQ, including the best algorithms to consider for each Life Cycle in the model. This version also continues to improve performance and speed, and enables additional predictive algorithms that expand the options that can be employed by Othot to solve customer HIQs.
Quality
This Othot platform release improves the overall quality of the platform by addressing several functional and usability issues reported by the Othot team and by customers. One major aspect of these improvements is the continued expansion of the Othot solution on the Google Cloud infrastructure. Othot continues to leverage the power and scalability of Google behind the scenes to ensure that Othot provides the most scalable, secure, and reliable solution on the market.
Next Release
The Othot development team has already begun working on the next release, 2.0. This release is scheduled to go into production in early 2018. For more details about the features planned for the 2.0 release, contact Othot Customer Success or Othot Sales.