Application Architect at Canada Border Services Agency
Real User
Top 20
2024-07-12T19:21:19Z
Jul 12, 2024
There are performance issues. Extracting data from many combined tables can take hours and occasionally crash the server due to memory leaks. This performance problem bothers people. The performance issue seems to be related to the server. We design streams on the client and submit them to the server, which generates a large SQL statement. There are two potential bottlenecks: one in the server and another in data extraction. I'm unsure about the exact mechanics of data splitting when fetching from the database. When streams become larger, performance bottlenecks may occur in the IBM SPSS Modeler server or the database. Sometimes the server crashes and needs to be restarted to release memory on both sides. I'm not sure exactly where the problem is caused, as I focus on stream design rather than server issues. The problem could be on the IBM SPSS Modeler server and database.
The platform's cloud version needs improvements. The process to access workflow could be user-friendly. It could be easier to log in and manage security levels. Additionally, it needs to be more popularized and introduce customization options for small companies along with enterprises.
I work with many data and tags. The product does not have a search function for tags. I cannot search for the name of the tag. I need to go through a list to find it. The software is old. The solution must be updated. It should also be a little bit faster. If we change the name of a field, the change should reflect in all the nodes we use later.
Principal Scientist I at a manufacturing company with 10,001+ employees
Real User
Top 10
2023-06-08T21:09:00Z
Jun 8, 2023
Speaking about the room for improvement, I can say the solution is outdated. It would be good if there was potentially some space in the solution for open-source languages like Python or other coding languages associated with analytics, PySpark, or Python.
It is not free to use. The forecasting could be a bit easier. They should include more recent techniques. A lot of competition is currently being built. IBM needs to increase its capabilities. Otherwise, it will be overtaken by other options. Sometimes IBM gives you the graphs only. There needs to be more visualizations for non-technical people.
Professor of Data Mining at Universidad Politecnica de Madrid
Real User
2022-05-09T16:51:28Z
May 9, 2022
Neural networks are quite simple, and now neural networks are evolving to these architecture related to deep learning, etc. They didn't incorporate this in IBM SPSS Modeler.
The time series should be improved. The time series is a very important issue, however, it is not given its value in the package as it should be. They have only maybe one or two nodes. It needs more than that. Also, it needs to be easier to use, for instance, you have, for the regression techniques, an assembled way for the automation that the model can detect the type of the logistic regression. If it is binary or multinomial or whatever. For the time field, they have an expert model, however, it is not as strong as regression techniques. Therefore, they need to work more on the time series. Right now, with the Modeler, using unstructured data means needing to pay attention to IBM Modeler, including how to deal with the pictures. Currently the data, in the beginning, was structured, like an organized spreadsheet, however, now, you have to use unstructured data like pictures, voice, even location maps. This area needs improvement. If they can add this to the Modeler, it would be number one around the world.
Time Series or forecasting needs to be easier. It is a very important feature, and it should be made easier and more automated to use. For instance, for logistic regression, binary or multinomial is used automatically based on the type of the target variable. I wish they can make Time Series easier to use in a similar way.
Contracts Manager at a program development consultancy with 1,001-5,000 employees
Real User
2020-10-11T08:58:11Z
Oct 11, 2020
I think that Modeler needs to be more commercially effective because, of the competing tools, some of them are free and others are available at a very nominal cost. When you are not using the product, such as during the pandemic where we had worldwide lockdowns, you still have to pay for the licensing. It is just wasting the term and they should have suspended the fees and extended the licensing timeline. Essentially, you can't stop it, even if you're not using it, and it is a little difficult to accept the cost in such situations. What we really need is some flexible terms with respect to the renewal or a break from the strict license timeline.
Graduate Teaching Assistant at a non-profit with 5,001-10,000 employees
Real User
2020-09-22T07:16:10Z
Sep 22, 2020
I actually think it is a great product. Maybe there could be some enhancement with more extensive help built into the interface. This could help end-users to understand the features as well as how to use them. Apart from that, it is a great product.
Dimension reduction is very important, especially if you are working with millions of recordings and thousands of variables. It exists already, but it should be classified separately. The solution could be improved by adding a feature for statistical analysis like processes. They have some in the output, but not in the modes itself. I hope they can add statistical knowledge to the solution.
Application Architect at a government with 10,001+ employees
Real User
2020-08-19T07:57:26Z
Aug 19, 2020
This is an expensive predicament software solution. Currently, the terminals offer the tools for the data analytics, but it needs development. There's a limit to the license. For the data analytics, it's very similar to Tableau. The solution has lots of branches, departments and the teams - that makes it quite a complex solution. We don't always use or need the major developmental version. We care about the KPI. We only care about the KPI reporting so it would be helpful if things were simplified.
Director - Institute of Advanced Analytics at a university with 1,001-5,000 employees
Real User
2018-08-06T08:33:00Z
Aug 6, 2018
I understand that it takes some time to incorporate some of the new algorithms that have come out in the last few months, in the literature. For example, there is an algorithm based on how ants search for food. And there are some algorithms that have now been developed to complement rules. So that's one of the things that we need to have incorporated into it.
Lecturer at School of Science, University of Phayao
Real User
2018-05-24T05:34:00Z
May 24, 2018
Data encoding is friendly for UTF-8. The unstructured data is not appropriate for SPSS Modeler. Finally, the standard package (personal) is not supported for database connection.
* Formula writing is not straightforward for an Excel user. Totally new set of functions, and it takes time to learn and teach. * Automating procedures: Writing macros is not easy and difficult to learn. * It is not integrated with Qlik, Tableau, and Power BI. Unfortunately… * Expensive to deploy solutions. You need to buy an extra deployment unit.
IBM SPSS Modeler is an extensive predictive analytics platform that is designed to bring predictive intelligence to decisions made by individuals, groups, systems and the enterprise. By providing a range of advanced algorithms and techniques that include text analytics, entity analytics, decision management and optimization, SPSS Modeler can help you consistently make the right decisions from the desktop or within operational...
There are performance issues. Extracting data from many combined tables can take hours and occasionally crash the server due to memory leaks. This performance problem bothers people. The performance issue seems to be related to the server. We design streams on the client and submit them to the server, which generates a large SQL statement. There are two potential bottlenecks: one in the server and another in data extraction. I'm unsure about the exact mechanics of data splitting when fetching from the database. When streams become larger, performance bottlenecks may occur in the IBM SPSS Modeler server or the database. Sometimes the server crashes and needs to be restarted to release memory on both sides. I'm not sure exactly where the problem is caused, as I focus on stream design rather than server issues. The problem could be on the IBM SPSS Modeler server and database.
The platform's cloud version needs improvements. The process to access workflow could be user-friendly. It could be easier to log in and manage security levels. Additionally, it needs to be more popularized and introduce customization options for small companies along with enterprises.
I work with many data and tags. The product does not have a search function for tags. I cannot search for the name of the tag. I need to go through a list to find it. The software is old. The solution must be updated. It should also be a little bit faster. If we change the name of a field, the change should reflect in all the nodes we use later.
The integration with sources and visualisation needs some improvement. The scalability needs improvement.
Speaking about the room for improvement, I can say the solution is outdated. It would be good if there was potentially some space in the solution for open-source languages like Python or other coding languages associated with analytics, PySpark, or Python.
It is not free to use. The forecasting could be a bit easier. They should include more recent techniques. A lot of competition is currently being built. IBM needs to increase its capabilities. Otherwise, it will be overtaken by other options. Sometimes IBM gives you the graphs only. There needs to be more visualizations for non-technical people.
Neural networks are quite simple, and now neural networks are evolving to these architecture related to deep learning, etc. They didn't incorporate this in IBM SPSS Modeler.
The time series should be improved. The time series is a very important issue, however, it is not given its value in the package as it should be. They have only maybe one or two nodes. It needs more than that. Also, it needs to be easier to use, for instance, you have, for the regression techniques, an assembled way for the automation that the model can detect the type of the logistic regression. If it is binary or multinomial or whatever. For the time field, they have an expert model, however, it is not as strong as regression techniques. Therefore, they need to work more on the time series. Right now, with the Modeler, using unstructured data means needing to pay attention to IBM Modeler, including how to deal with the pictures. Currently the data, in the beginning, was structured, like an organized spreadsheet, however, now, you have to use unstructured data like pictures, voice, even location maps. This area needs improvement. If they can add this to the Modeler, it would be number one around the world.
Time Series or forecasting needs to be easier. It is a very important feature, and it should be made easier and more automated to use. For instance, for logistic regression, binary or multinomial is used automatically based on the type of the target variable. I wish they can make Time Series easier to use in a similar way.
SAS type free facility to .edu email - free training by IBM skill training. Currently IBM charge USD 2,400.
I think that Modeler needs to be more commercially effective because, of the competing tools, some of them are free and others are available at a very nominal cost. When you are not using the product, such as during the pandemic where we had worldwide lockdowns, you still have to pay for the licensing. It is just wasting the term and they should have suspended the fees and extended the licensing timeline. Essentially, you can't stop it, even if you're not using it, and it is a little difficult to accept the cost in such situations. What we really need is some flexible terms with respect to the renewal or a break from the strict license timeline.
I actually think it is a great product. Maybe there could be some enhancement with more extensive help built into the interface. This could help end-users to understand the features as well as how to use them. Apart from that, it is a great product.
Dimension reduction is very important, especially if you are working with millions of recordings and thousands of variables. It exists already, but it should be classified separately. The solution could be improved by adding a feature for statistical analysis like processes. They have some in the output, but not in the modes itself. I hope they can add statistical knowledge to the solution.
This is an expensive predicament software solution. Currently, the terminals offer the tools for the data analytics, but it needs development. There's a limit to the license. For the data analytics, it's very similar to Tableau. The solution has lots of branches, departments and the teams - that makes it quite a complex solution. We don't always use or need the major developmental version. We care about the KPI. We only care about the KPI reporting so it would be helpful if things were simplified.
Weak documentation and user guide.
I understand that it takes some time to incorporate some of the new algorithms that have come out in the last few months, in the literature. For example, there is an algorithm based on how ants search for food. And there are some algorithms that have now been developed to complement rules. So that's one of the things that we need to have incorporated into it.
Data encoding is friendly for UTF-8. The unstructured data is not appropriate for SPSS Modeler. Finally, the standard package (personal) is not supported for database connection.
* Formula writing is not straightforward for an Excel user. Totally new set of functions, and it takes time to learn and teach. * Automating procedures: Writing macros is not easy and difficult to learn. * It is not integrated with Qlik, Tableau, and Power BI. Unfortunately… * Expensive to deploy solutions. You need to buy an extra deployment unit.