Content
Summary
COR AI Features
Technical details
Natural Language Processing
Machine Learning Cloud Service.
Summary
COR automates timesheets, through AI, to predict profitability, reduce work overload and improve client engagements.
The following document describes the technical specifications for COR AI usage into different features that bring meaningful solutions for end users.
COR AI Features
The list below enumerates the COR AI features available and provides a summary of the impact of each one:
|
Feature |
Notes |
|
Hour suggestions |
Algorithms that predict the number of hours worked per day in each task according to the behavior of the users who interact with them. This suggestion is unique per user, it adapts and depends on the usage pattern of each user. We currently have 4 different algorithms that compete with each other to provide the best suggestion according to the level of reliability based on the relationship between variables. |
|
User capacity |
Algorithm that predicts the level of occupation of the collaborators assigned to the tasks in a requested period of time, achieving a daily understanding of the effort of the actors involved, giving future visibility. |
|
Task categorizer |
Predictive text algorithm that allows the automatic categorization of tasks, according to their title, description, assigned project, collaborator tags, and notes described in tasks. With this algorithm, we provide users with predictions of the type and category to which the task belongs. The task categories are used by the platform at different levels to provide precision in time suggestions. |
|
Hour task estimate |
Algorithms that predict the estimated completion time of a task based on the historical intervention of those involved. We currently maintain 4 types of algorithms that compete with each other at different levels from macro to micro, such as calculations by segment, company, client, and projects which change over time depending on the characteristics of the tasks and how they evolve. |
|
Language detection |
Algorithm that can detect up to 100 different languages in a given text. Used to perform task categorization, it defines the language and allows the extension of the scope of training and identification of the model for its prediction. |
|
Project deviation |
Algorithm that measures the elapsed time between the end date set by the user and the true end of a project, being able to show the delay the project will have to be delivered and the real costs of the project. |
|
Profitability prediction |
Algorithm that predicts the current and future profitability of a project. Multiple analyses allow determining if a project or client will contribute or subtract value to the total profitability. |
|
Project Insights |
Algorithm for calculating the most profitable and efficient projects in an average of 20% of all the company's projects. Being able to have results like: - Less Profitable - Most Efficient - Less Efficient |
Technical details
Linear models
One of the most common models covered in COR, linear models simply describe the relationship between a set of input numbers and a set of output numbers through a linear function that lets our systems give Insights of projects and tasks.
Classification tasks often use a strongly related logistic model, which adds an additional transformation mapping the output of the linear function to the range [0, 1], interpreted as the “probability of being in the target class.” Linear models are fast to train and give a great baseline against which to compare more complex models.
Scikit-learn
Scikit-learn is a Python package designed to facilitate the use of machine learning and AI algorithms. Sci-kit learn includes algorithms used for classification, regression, and clustering. Scikit-learn includes useful tools to facilitate the use of machine learning algorithms.
Supervised Learning Models
Supervised learning, also known as supervised machine learning, is a subcategory of machine learning and artificial intelligence. It is defined by its use of labeled datasets to train algorithms that classify data or predict outcomes accurately. As input data is fed into the model, it adjusts its weights until the model has been fitted appropriately, which occurs as part of the cross-validation process.
COR uses different Supervised Learning Models in order to get tasks classified.
Custom Classification
COR Task Classification uses text classification models using each company's specific labels. Tasks are categorized through COR AI and suggest classification in real-time.
Natural Language Processing
Natural Language Processing (NLP) is a subfield of machine learning that makes it possible for computers to understand, analyze, manipulate and generate human language.
COR uses different NLP models in order to feed supervised learning models for task classification.
Keyphrase Extraction
A key phrase is a string containing a noun phrase that describes a particular thing. It generally consists of a noun and the modifiers that distinguish it. For example, "Design" is a noun; "a great banner" is a noun phrase that includes an article ("a") and an adjective ("great"). Each key phrase includes a score that indicates the level of confidence that COR has that the string is a noun phrase. COR AI Algorithm uses the score to determine if the detection has high enough confidence for further classification.
Language detection
COR AI examines text to determine the dominant language inside a task. COR AI identifies the language using identifiers from RFC 5646 — if there is a 2-letter ISO 639-1 identifier, with a regional subtag if necessary, it uses that. Otherwise, it uses the ISO 639-2 3-letter code. For more information about RFC 5646, see Tags for identifying languages on the IETF Tools website.
Machine Learning Cloud Service
It provides common machine learning algorithms that are optimized to run extremely large data in a distributed environment. COR deployed models into a secure and scalable environment, to execute different algorithms, for access to data sources for exploration and analysis.
The platform automates the work of building a ready AI pipeline and simplifies the process to use common algorithms and other tools in the machine learning process.
Data training teaches machine learning to behave in a certain way based on recurring pattern recognition within data sets. The data is then inferred on how to respond to new data patterns to be served as new features.