Modelcraft | Management Solutions

Why ModelCraft™?

ModelCraft™ is an automated machine learning and component-based modelling tool that allows simplifying and industrializing the development and validation of advanced analytics models

ModelCraft™ has intuitive components for data processing, advanced analytics model development and interpretability. It allows users to connect to any data repository and to generate modelling workflows in a visual manner and without the need for programming.

ModelCraft™ thus responds to industry needs such as:

Incorporating advanced machine learning algorithms into the modelling process in a simple and intuitive way, such as gradient boosting, elastic nets, ensemble models, random forests, support vector machines or neural networks, among others.

Industrialize and accelerate the modelling process, reuse codes and reduce the time spent on routine and repetitive tasks.

Facilitate access to advanced modelling techniques for non-specialist teams.

Improve the performance, traceability and comparability of models.

Enable the interpretability of models and automate the generation of their documentation.

In short, to improve the model development and validation processes, and to enable the generation of advanced challenger models.

In addition, ModelCraft™ has been developed in the cloud and natively uses the capabilities of auto-scalability and parallel computing.

Thanks to this, ModelCraft™ is a tool to accelerate the transition to a component-based modelling, which is more efficient, more interpretable, more accessible and with lower model risk.

Functionalities of ModelCraft™

ModelCraft™ has a library of predefined components that allows developing the complete modelling lifecycle

Input, connection and download

Input data reader with different formats and delimiters

Model reader in .pkl format

Database connection components (cloud, RDS)

Download of models, files and reports

Sampling

Sampling algorithms

Obtaining training, test and validation samples

Algorithms for processing highly unbalanced data

Visualizations

Model comparison and selection of the optimal model

Exploratory data analysis

Representation of histograms, pie charts, etc.

Analysis of database structure

Calculation of descriptive statistics

Performance and prediction

Mapping of criteria to training samples, validation and testing

Calculation of metrics and KPIs

Out-of-time analysis and backtesting

Predictions

Model validation

Database management

Simultaneous connection to different data repositories in various languages (postgresql, mariadb, mysql, oracle, terminal, etc.)

SQL data processing (generation of tables, data manipulation and fetching, table traversal, etc.)

Import and export of files

Feature engineering

Binning algorithms

Correlation analysis

Variable selection and elimination

Missing variables analysis and imputation

Outlier analysis and treatment

Cálculo de weight of evidence.

Encoding techniques

Model training

Hyperparameter search techniques

Ensemble methodologies

Genetic algorithms

Training of supervised and unsupervised models

Neural networks

Interpretability

Variable importance analysis

Branch & bound algorithms

Surrogate model building

Univariate and bivariate analyses

Interpretable trace of the entire development process

Automatic generation of a complete model report

Automated machine learning

Generation of user-created components, including free code, in real time

Determination and optimization of the critical path between components

Encapsulation of components in three levels of depth

Access to Python and R code generated by each component

Features of ModelCraft™

ModelCraft™ enables the application of machine learning techniques in an intuitive and robust way, and it is deployed in a simple, flexible and scalable technology

User

Ease of use:

The ModelCraft™ interface allows the visual buildup of modelling flows and gives access to all parameterization in an organized way.

Personalization

Users can create their own components, design their own modelling flows and access the generated Python and R codes. .

Explainability

ModelCraft™ can generate an automatic summary of the model, its variables and its complete development process at the click of a button, describing the decisions made by the user (e.g. the model's hyperparameters) and by the algorithms (e.g. the exclusion rationale for each variable).

Evolution

ModelCraft™ is constantly evolving and adapting to the needs of the industry; for example, the probability of default (PD) calculation has already been added to the ModelCraft™ component library, and LGD and CCF will be added shortly.

Traceability

The interface offers full traceability of the developments made, allows access to all the code, and provides logs and traceability functionalities.

Optimized code

Each component in ModelCraft™ has been optimized to maximize computational power, using libraries available from the most advanced programming environments.

Integration

ModelCraft™ integrates natively with two proprietary Management Solutions: ModelCraft Integrator™, for putting models into production through a model and rule manager that interacts with operational systems, and ModelCraft Monitor™, for periodic monitoring and backtesting of models.

Deployment flexibility

ModelCraft™ puede implementarse como herramienta on-premise o como software-as-a-service (SaaS) en la nube, de acuerdo con las necesidades del cliente.

Scalability and parallel computing

ModelCraft™ is scalable and takes full advantage of parallel computing capabilities, allowing the processing of large volumes of data and executions with a high computational load.

Easy deployment

The designed architecture allows ModelCraft™ to be deployed in a very short time, including an analysis of customer needs, their configuration preferences and necessary customizations. .

Reduced support

ModelCraft™ requires reduced IT support, and maintenance includes the deployment of new releases.

Structure of ModelCraft™

ModelCraft™ has two work areas: the data management panel and the modelling panel

The data management panel allows users to connect to any database, and extract, merge and load tables with SQL language. The modelling panel allows, through components, to perform exploratory data analysis, prepare data for modelling, and train and interpret advanced machine learning algorithms.

Data management

Database management
Simultaneous connection to different data repositories in different languages (postgresql, mariadb, mysql, oracle, terminal, etc.)
SQL programming

Modelling

Data processing and EDA
Modelling components and advanced machine learning algorithms
Interpretability and automatic reporting components

In turn, the modelling components are structured in three levels of depth: from level 0 (minimum executable task, such as outliers detection or model training) to level 2 (aggregation of components to build an end-to-end modelling process)

Modelcraft Structure

IT architecture

ModelCraft™ has been developed cloud-natively, using a Kubernetes-based architecture that enables rapid deployment of the component library and a visual programming frontend.

The ModelCraft™ architecture is focused on different stacked component libraries, allowing the use of the components as microservices, and includes a user, storage and execution manager.

Credentials

Management Solutions has extensive capabilities and references in the field of advanced modelling, combined with in-depth knowledge of technology and business processes

Management Solutions provides differential value in the field of advanced modeling:

It has extensive experience in the development, validation and implementation of models and algorithms applied to multiple business areas in different industries and sectors.
It has quantitative capabilities and resources with a high degree of knowledge and specialization (two thirds of its staff are STEM).
Specifically, it has teams of data scientists who carry out projects in artificial intelligence and machine learning, as well as an R&D team dedicated to the practical application of AI applied to management.
It has developed a Chair in Big Data and Analytics together with the Universidad Politécnica de Madrid, where it develops lines of research on modelling techniques and artificial intelligence. This has allowed it to develop ModelCraft™ and keep it up to date with the latest advances in modelling and computing.

Thanks to these differential values, Management Solutions implements ModelCraft™ in all its industries and in all the countries where it operates.

Financial entities

Energy & Utilities

Telecommunications

Other sectors: manufacturing, pharma, logistics and transportation, tourism, professional services, education, construction and infrastructure, government