Reviews (1-25 of 48)
- Integration of IBM Watson APIs such as speech to text, image recognition, personality insights, etc.
- SPSS modeler and neural network model provide no-code environments for data scientists to build pipelines quickly.
- Enforced best-practices set up POCs for deployment in production with a minimum of re-work.
- Estimator validation lets data scientists test and prove different models.
- Watson Studio's UI is not always intuitive, especially when it comes to requirements and specific settings.
- Documentation is not strong; tutorials and walkthroughs are noticeably light.
- Tight integration with IBM APIs also means less well-made integrations to third party data sources and APIs—MySQL support notably absent.
- It is relatively easy to use
- It works seamlessly with multiple languages.
- Its administration is surprisingly easy
- And it's easy to install / upgrade / maintain
- Need better training materials for data scientists. Especially the ones who are not formally educated as data scientists.
- The videos in the tutorials are all on Youtube which are usually blocked on most work campuses.
- And the IBM Think campus training could have been better as well
Less suited for lower level data analysis which does not add much value.
- Intuitive GUI for us to begin using the studio
- It works well that we can embed the decisions into our existing offering without a lot of changes
- The pricing model is flexible
- Like the opportunity to embed more data
- Provide hint to use services based on verticals
- Suggest how it could be embedded into mobile apps
- Would like to understand the deployment model better
1. Offering ancillary upgrades to airline passengers
2. Predicting flight delays based on historical patterns coupled with live feeds like weather
- It is easy to use and I don't need to have a team full of data scientists to use them
- It is easy to deploy when the models are trained and we don't need to hire many software engineers to take care of deployment
- It allows us to test different models rapidly and so helps to accelerate the product development process
- The cost is steep and so only companies with resources can afford it
- It will be nice to have Chinese versions so that Chinese engineers can also use it easily
- It takes a while to learn how to input different kinds of skin defects for detection
- Ease of use and quick to explore
- Guided experiences and ability to leverage multiple algorithms to identify the best one
- Great support and sales teams
- There isn’t much I think I can provide critical or improvement feedback on
- Feature rich. IBM DSx provides a plethora of tools to leverage the use of data science in your organization and suit your specific needs.
- IBM DSx supports a huge variety of sources of data. From your traditional SQL database to every major data warehouse, IBM DSx does a great job at connecting to or pulling from your data source.
- Its greatest strength is the fact that is a cloud-based service. There's no need to waste time on configuring and maintaining an environment to start analyzing data, which may not be an easy task.
- Pricing. The price for this product is quite steep and, since it features so many solutions, it makes sense to cost as much as it does. But the creation of personal plans with fewer features might prove interesting to bring the product to a broader audience, like enthusiasts that are starting to get in touch with data science.
- Some issues regarding notebooks and the use of data refinery are quite annoying to the experience because, depending on the use that you make of it, they might appear quite regularly.
- Lack of a changelog. Like many IBM products and platforms, DSx is in constant development and is updated regularly. This is a great point, except for the fact that sometimes it lacks a changelog to properly inform what has been changed, requiring the user to investigate on its own.
- Cloud-based file sharing helped our organization stay up to date when managing assets, new or old.
- Watson studio does a fantastic job visualizing outcome data which enabled our organization to easily create a narrative based on what we were able to see.
- Particularly within our organization, Watson Studio strength was noticed in its ability to processes enormous amounts of data in such a short amount of time.
- Watson Studio could used improvement in its user-based community. I'd like to see more local and remote events showcasing its potential.
- Watson Studio could improve by providing its users, use-cases that leverage data in unusual ways.
- We think Watson studio could also improve by decreasing its price in order to capture new talent in the data industry.
When I do get a response, it doesn't solve my problems. IBM needs to get it's act together or risk losing out.
Using R and Python interfaces on the same platform has great flexibility.
Drop down menus are great! Nothing needs changing there!
Apache Spark connection seems to be unstable.
The H2o package in R is erratic.
Apache spark in R could use some documentation - solid documentation.
- You can use SPSS model in order to predict trend with historical data
- You can use R in order to clean up your data a Jupyter notebook
- You can use Jupyter Notebooks to analyze Twitter data and create data visualizations
- We try to install DSX in the local environment but it needs more resources
- I'd like a better visualization library for charts
- I'd like more webinar in order to introduce to the platform, also in Italian language
- DSx is particularly well suited for ML data prep. It's easy to ingest from many different kinds of sources and then perform various cleansing, transformation, and enrichment operations.
- DSx makes collaboration with other team members very easy. Control over who can see and interact with each project is straightforward and simple to administer.
- DSx doesn't create proprietary lock-in. Notebooks can be exported in a number of different forms to share people that don't use DSx or to run in a different environment.
- Stability has gradually improved over the past year or so, but could still be better.
- I'd like to see options for leveraging a GPU on the cloud-hosted version.
- I'd like to see even more ML model lifecycle support, but it's my understanding that this is coming with the move to Watson Studio.
- To have Jupyter notebooks and RStudio in the same environment is great!
- The free Spark engine is perfect and enough to support the development activities.
- The integration with GitHub facilitated our collaborative work.
- To set up a new Spark cluster and use it with DSx is a bit hard. It would be great to have the option to create a new big cluster without leaving DSx.
- I've faced several problems with DSx desktop.
- Configuration: You can forget about all the setup. You just open a notebook, import the libraries you want and start writing
- Sharing: You can share your thoughts with anyone, because all your code lives in the cloud
- Tools: IBM has amazing tools for speech recognition, image processing, and so on
- Because of my use, I didn't find anything to improve. I think that making things more visual will be useful for non-expert people, like flowcharts for example
After that, I had to implement this model to be used as a REST service. I tried to do that with DSX but it was not possible, which is reasonable since it is not designed to do that.
- I like very much to have a multitude of tools integrated nicely in one place.
- The documentation is ever-growing and a source of continuous new insights. I appreciate the notable curatorial effort.
- The tools and services are ever-growing too. The focus on open-source tools and the "no barriers in the middle" is really one of the strongest points in my opinion.
- Well, I had issues of performance and responsivity.
- As a tool for exploring new ideas and learning new techniques the free plan is limited to very small datasets in order to perform in a reasonable time.
- As an example a simple test of Random Forest on a dataset of 100k rows didn't finish in the RStudio provided by Watson whereas it took half an hour in the RStudio installed in my desktop.
- The run period is good and what you run is saved.
- You have to install nothing and that's a great advantage.
- You can create conections to save your files in the cloud.
- It works perfectly in a collaborative environment and it's easy.
- It is limited still. Jupyter supports many programming languages but DSx just manages three.
- The scheduled jobs are fantastic but it can be more frequent.
- The multiple errors of the account, the group, the services are confusing.
- Flexibility: all the other solutions we procure were modular with black boxes.
- Data center is in Europe, since in my company we can't have data in US
- From all the solutions we procure IBM was the only team that actually embarked on the project and didn't only tried to sell a tool/service.
- The download of files is terrible
- Managing the files is terrible
- Data ingestion, using Data Refinery and Cloud Object Storage
- Data persistence, using Db2 Warehouse on Cloud
- Data manipulation, using Jupyter notebooks, via R or Python scripts
- Data visualisation, using Pixie dust or Watson Analytics
- No ability to audit/log activity
- Difficult to secure
- No access to underlying infrastructure
- Poor reliability and performance
- Machine learning capability too basic/simple and black box nature means it is difficult to validate/trust
- Limited ability to customise Watson Analytics visualisation
- Poor support response times
- Poor native support for Softlayer S3 storage
- Allows for people with various technology backgrounds to use a common platform.
- Easy implementation due to the cloud availability.
- The DSX platform allows for junior and citizen data scientists to perform complex actions without needing to have deep knowledge of some of the underlying configuration and setup that generally come with standalone/local analytics tools.
- Interfacing with non-IBM technologies is often cumbersome and sometimes restrictive.
- The interface has undergone a number facelifts which often causes some lost productivity when you need to "find" where things have moved to.
Moving the models developed to a production ready model is not an easy path and often taking the analytics idea to a product involves translating the method and approach to other tools.
- DSx provides an excellent support for machine learning modeling
- DSx provides a good environment for collaboration between colleagues
- DSx also provides support for sharing datasets, models, notebooks, and articles to start projects
- They should involve the drag and drop functionality more into DSx for data analysts who are not so much into coding
- Also, scripting nodes should be integrated into the drag and drop(SPSS MODELER)
- Also, more nodes should be added to the SPSS Modeler. For example, remove duplicates node, edit metadata node etc.
DSX is fulfilling the capabilities that all users need. We see the main core of users on people who want to experiment with analytics, algorithms, and techniques before going full business implementation.
We used the DSx platform in the context of a data science project in the medical domain. The general problem was to predict the health condition of a patient in real-time for the upcoming minutes based on various features that were provided by the customer. The status of a patient could be described by a limited number of classes which allowed us to interpret it as a classification problem.
Due to confidentiality reasons, we had to perform all tasks on the DSx. This included the analysis of the data set, the computation of additional features, the development and optimization of a machine learning model, as well as the analysis of the results. Therefore, we relied in particular on Jypiter notebooks (python and R) and RStudio
- Standard software packages (python and R) are available and ready to run.
- Data from various sources (e.g. external databases) accessed and loaded from DSx.
- Customer support provides valuable guidance and helps to solve problems.
- An actual IDE for python would be very helpful.
- Some python packages were not up-to-date and it was not possible to install the current version.
- It should be somehow possible to monitor the used resources and system load (CPU/RAM).
IBM Watson Studio Scorecard Summary
Feature Scorecard Summary
About IBM Watson Studio
IBM Watson Studio Technical Details