Question 1

Does Vantage facilitate to import (open source) models (i.e. associated with auto/weather, etc.)? Thus, I train it on existing data residing in Teradata (cloud) and further evaluate/score for better prediction. If yes, what is the name of this tool?

Accepted Answer

Yes, it does. The tool is Teradata Vantage Analytics Platform. With Vantage, you can create your own models with native functions (e.g. XGBoost, GLM) and score them natively as well. Alternatively, you can also import open source models that are generated through open source Python and R packages (e.g. scikitlearn). These models, which are already trained externally, can be brought into Vantage to be scored on all the data.

Question 2

Does the data have to be in Teradata Vantage in order to use these analytic tools?

Accepted Answer

Not at all. Connectors within Vantage allow bi-directional movement of data and analytics. You don´t need to have the data stored in or copied into Vantage. For example, you can connect an S3 bucket or a Hadoop data lake through QueryGrid and access the data to run analytics. The value of using Vantage is that the analytics is led to the data and not the other way around.

Question 3

Are there built-in governance options such as data sheets/catalogues and model cards?

Accepted Answer

At Teradata, we work with partners in the field of governance, data cataloging and data collection. For example, Alation, whose solution integrates with Vantage. As part of Vantage, you can look at the lineage of your data set that you are running your models on, which are being catalogued within the platform itself.

Question 4

World clouds are a common method. What is unique about your tools?

Accepted Answer

Typically, to create the data visualization of a world cloud, different kinds of analytics have to be performed – e.g. text parsing or set certain values for engrams to see which words are close to each other or how large a data set is in one screen. All the different steps that need to be done for a word cloud would each require a different tool with separate steps and data preparation in the open source environment.

Teradata Vantage allows an end-to-end analytical process and environment: You access the data, you run the analytics functions – regardless of how large the data is in one screen. For example, the sentiment extractor function in Vantage allows you to get the positives and the negatives; and every single word can be evaluated based on whether it's positive, negative or neutral because it's supervised learning and there's a dictionary that runs on it. The word scores get put together in one particular phrase, and then get visualized using a frequency-based approach. All this happens on exactly one platform – Teradata Vantage – without you needing to do any of these things separately, which is a significant advantage, not only in terms of the amount of time that you save but also in terms of how many different people in the organization you can empower, to be able to do all these kinds of analytics right without people having to actually code and gram splitter etc.

Question 5

Does Vantage offer dynamic of visualization of geolocation analytics?

Accepted Answer

Yes, geospatial analysis is an essential part of Vantage and visualizations are provided natively, as well as through our partners like Tableau or MicroStrategy. For example, within Vantage you can take location data as it changes over time and visualize it on a map. Also, using the connectors into Tableau, you are able to transfer all the data in order to have additional visualizations.

Question 6

Are you using API as a connector?

Accepted Answer

APIs are used to be able to extract data from sources for which there is no native connector (e.g., Twitter data).

Question 7

Bring-your-own-models: If I build my models in Python, how do I interact with Vantage and run the model on Vantage?

Accepted Answer

Yes, you can bring external models and transform them to the scheme of Vantage and run the operationalization. E.g. You can run a XGBoost model with packages in Python, R and SQL on Vantage, as long as that model is transformed to a specific schema/structure of the model that the Vantage operationalization function expects. Vantage delivers a lot of flexibility to either use the models and algorithms within Vantage or to craft your own algorithms in your choice language.

Question 8

Does Teradata Vantage provide a pool of compute on demand, similar to what cloud companies provide, e.g. Synapse provides an SQL or Spark pool?

Accepted Answer

Yes. Vantage systems, available on the public and private clouds (AWS, GCP, Azure), as well as-a-service and on-premises, are platforms that can be increased for both capacity and compute based on the analytic needs of the customers.

Question 9

Can I use the advanced analytic functions within Vantage with existing BI tools?

Accepted Answer

Absolutely. In fact, through our ODBC connectivity, you can use any of the BI tools such as Tableau, MSTR, Cognos, QLIK and others. We partner with the best of breed BI companies.

Question 10

Can sentiment analysis parse emoticons?

Accepted Answer

Yes. We create a custom lexicon which includes emoticons and their corresponding sentiments. If needed, new emoticons can be added to the lexicon with their corresponding sentiments and these can then be parsed through the native sentiment extractor function on ALL the text raw data to get the overall sentiments.

Question 11

What are the best practices for analyzing data sets that span different localities, e.g. cloud native data vs. Data center native data?

Accepted Answer

All data is accessed via Vantage (not always ingested into it though as users may want data to stay in their source systems) and regardless of where the data is loaded Vantage ensures that there is a common identifier to link the data across different locations. Each location has a unique identifier to indicate its source and each higher-level record (e.g. customer name) is mapped to a unique identifier that spans these locations. This way, we can easily analyze the data at the customer level or the location level or any other level based on the granular identifiers that are applied. This is a very common occurrence in today’s world of hybrid data storage and Teradata has had numerous applications of this with our customers.

Question 12

Teradata has always been known for the performance it brings using parallelism. Are there any apprehensions to be watchful when running advanced analytics?

Accepted Answer

The same parallelism which brought us scaled performance is brought to the analytics enterprise. In fact, ALL of the architectural considerations for Vantage took maximal advantage of all the prior work on parallelism that was invested in the EDW world. Vantage has retained all these core elements that have been brought to bear in doing enterprise wide analytics across a variety of use cases for numerous customers. In other words, no apprehensions at all!

Question 13

Can I have multiple outcome sequences when doing path analysis?

Accepted Answer

Yes, absolutely. In fact, not only can you have multiple patterns that you are investigating, you can also have several wild cards that can be included. There is a whole slew of such wild cards and pattern qualifiers that are present in the nPath syntax that can be used.

ANSWERS

Vantage Interactive Webinar

Most Frequently Asked Questions From This Session: