BDB Predictive Workbench

Augment Business Decisions using Predictive Analytics

Empower frontline decision makers with valuable insights by using predictive models and machine learning. Analyze the factors behind each move to decide the next probable business strategy.

What is Predictive Workbench?

Predictive Workbench comes with Integrated Algorithms from R, Spark ML, Python, Keras + Tensorflow to create workflows and derive business insights. Custom Algorithms in R, Spark ML (Scala), Python can be designed and utilised as and when required. Sentiment Analytics, Image and Video analytics can also be performed using Deep Learning Workbench.

The Predictive Analytics workbench

								
Connect to a Data SourceDetermine what you want to predict or
									understand. Define the aspect on which you want to use BDB
									Predictive Workbench.

Data Preparation – Transform, cleanse and format the data setAcquire data in any form and modify
									to make it analysis ready. Use inbuilt allied components to
									prepare and transform data.

Prediction BuildingSplit data into train/test, drag & drop components in a web UI, configure algorithms, execute, view 
								summary & results , compare their performance, save the workflow or model and deploy as a service.

Captivating VisualizationDisplay the predicted results through
									rich and interactive visuals for well-informed insight.

Watch BDB Predictive Workbench in Action.

Forecast Model

Creation of Forecast Model in BDB Neural Network Workspace

RNN Model

Creation of RNN model in BDB Neural Network workspace

Python Workbench

Weather Prediction using Regression

3.5 Release

New features which have been released in 3.5 release.

HR Advanced Analytics

Predict the salary of an employee based on his/her skill and experience.

Spark ML recommendation engine

Product recommendation using Spark ML – ALS

Credit card scoring

Predict credit worthiness/score for loan applicants

Capabilities

								
Visualize Market Trends & Take ActionsUsing our regression analysis capabilities like Linear Regression, Multiple Linear Regression you can instantly predict market trends and foresee changes in demand and supply.

Custom ScriptsCustom Scripts lets you write down your R/ Spark Scala/ Python / Py Spark scripts. Custom Scripts lets you create customized algorithms for R, Spark, and Python environments.

Classification TechniquesUse our classification techniques like Decision Tree or Decision Matrix, Naive Bayes, Pareto Analysis, Conjoint Analysis (Via custom R) to make optimized and well-informed decisions.

Get Relational Insights Uncover invisible patterns and associations to glean predictive insights into your data using our Market Basket Analysis. 

Predictive ModellingHighly scaled (in performance) Predictive Models based on Decision Tree, clustering, Logistic Regression can be created for accurate prediction. 

Deploy Models in Machine Learning/ Data PipelinesDeploy the trained predictive models in ML-based Data Pipeline and combine them with other test datasets to produce seamless insights. Visualize the predictive results through rich and interactive dashboards.

How to use Predictive Workbench?

Experience the most intuitive and interactive predictive models. BDB Predictive Workbench, a web based UI, brings together advanced analytics spanning ad-hoc statistical analysis, predictive modeling, data mining, text analytics, entity analytics, optimization, real-time scoring, and machine learning.

Salient features of Predictive Analytics

Workbench User Interface - Drag and drop panel for the user to easily configure and build a predictive workflow.
Component Properties - Configure the independent/dependent columns, features/labels and allied algorithm properties through the Component Tab.

Console – Keeps you updated on the progress of the workflow at real time.

Summary – Key statistical information provided by the algorithms are displayed here, once the algorithms complete execution.

Results – view the predicted output in a data grid.

Visualization – Visualize the results using Bizviz charts. Integration with third party charts are also supported – RBokeh and Matplotlib
Data Source Supported - It supports most of the key data sources including CSV, RDBMS, Cassandra, Elastic Search, Spark SQL .
Data Preparation - filters, basic transformation, normalization, split, formula, data type definition, missing value replacement, sampling. Specifically for Spark - String indexer, Rformula, PCA, Chi Square, index to string, Spark SQL transformer, group by.
Save and reuse workflows - once you are convinced with the workflows, you can go ahead and save it for later use.
Save only the models - Provision to only save the spark/R/Python models (.rda, spark model) for reuse in machine learning pipelines.

Supported Algorithms -
Keras + Tensorflow – Train/Retrain Neural Network, Pre trained Sentiment Analysis model,Text Classification Model and Object Detection MOdel.

R - Clustering – Kmeans, Forecasting – Single/Double/Triple/ARIMA, Auto Forecasting, Market Basket Analysis, Regression - Linear/Multiple/Logistic, Outliers – inter quartile range, Classification – CNR/NB, Correlation, XG Boosting.

Spark - Clustering, Classification – NB, Decision trees, Random Forest, Recommendation engine using Alternate least square (ALS)

Python - Regression – linear/multiple/ logistic; Classification – Random Forest/Naïve-Bayes/ XG Boost
Create R/Spark Scala/Python scripts - Furthermore, the Custom R/Scala/Python Script component facilitates the user to script custom algorithms and functionalities.
Data Writer - functionality to save the predicted data into an RDBMS/Cassandra
Share the workflows/scripts with other Bizviz Platform users/groups - Saved workflows and scripts can be shared with other users and groups with appropriate permission levels.
Scheduling - Provision to schedule the workflow on a timely basis.