Notifications
Back

Unique step-oriented features available only in DAVinCI LABS

SPECIAL FEATURES

BY WORK PROCESS

Data Preprocessing


Easy data preprocessing such as missing values, outlier, and normalization

Constantly organizing data in a busy business environment is an extremely demanding process. Normally, data contains missing value and outlier.
DAVinCI LABS provides special feature that transform the data into a suitable format for machine learning model.

Data visualization function to get full picture of data

distribution and data characteristics at a glance

Feature Selection


Auto Feature Selection + Auto Feature Generation

that enhances modeling performance

Feature Type SwitchEasily switch the type of features, from numerical, categorical to letter-type.

Auto Feature SelectionPresent the most efficient-learning option that optimizes the model performance through full testing of all features.

Auto Feature EngineeringOnce you determine the preferred number of derivative variables and the number of variable you want to utilize, DAVinCI LABS automatically provides the optimum number of derivatives variables performance-wise and calculate its importance.

Are you testing each variable to figure out the optimal one?
Would you like to further improve your model performance by combining the already existing variables?
Leave your variable-related concerns to DAVinCI LABS.

Algorithm Selection


Optimal combination of algorithms that

precisely reflects your data characteristics 

The world’s only stabilized deep-learning algorithm reflected
Implements nearly infinite algorithm combinations through Ensemble and Cascade features
Automatically recommend algorithm that is most suitable to your data
 

Parameter Tuning


Various options to enhance model performance

Detailed parameter tuning options to improve the performance of each algorithm
You can use auto-tuning option to automatically get the optimized parameter, or manually tune the parameter and compare the performance

Automatic parameter tuning
Advanced parameter tuning

 

Performance Indicators allows easy comparison of algorithm performances

Check the list of algorithms ranked by performance as well as automatically tuned parameter values.


The following indicators are also provided to offer detailed analysis of respective algorithm performance.


Variable Importance

Distribution chart (comparison between actual and predicted value)

Confusion Matrix

AUPRC

AUROC

Decision Tree 

Model Result Analysis


Explanation & Report that provide detailed information on the model

Every machine-learning scientist faces difficulty in trying to clearly explain the correlation between the generated model and the data.

In order to solve this “blackbox” problem, DAVinCI LABS provides predictive explanation that clearly presents the reliability and influences of the inputted variables to the model.

DAVinCI LABS also provides a detailed report option that allows user to compare all performance index of training and validation data at a glance.

Prediction Explanation
Detailed report

Model Operation and Update


Deploy the model within one-click

DAVinCI LABS generated models can be easily deployed to your system.


With a single click, the model can be deployed via prediction server or downloaded as programming language


When the data is continuously aggregated, you can schedule auto model update cycle, or set a notification to manually update the model.


-Download prediction model

-Deploy REST API based prediction server

-Automatic update

 

Automation feature on every step

Easy automation / Manual enhancement

Select freely according to your needs

Easily get modeling result via

SHORT CUT MODELING


Experience the analysis results in  mere three clicks through the baseline model,

where everything from data preprocessing to parameter tuning, and algorithm selection is automated.

Generate and deploy advanced model via 

ADVANCED MODELING


DAVinCI LABS offers both full automation and manual tuning for model development.

Users can directly intervene in model development process to get an advanced model.


Target selection &

Data distribution check

Data

Preprocessing

Algorithm selection &

Parameter tuning