Press "Enter" to skip to content

Author: Kevin Feasel

Contrasting Three Classification Algorithms for Small Datasets

Published 2025-08-26 by Kevin Feasel

Jayita Gulati compares a few mechanisms to classify data:

When you have a small dataset, choosing the right machine learning model can make a big difference. Three popular options are logistic regression, support vector machines (SVMs), and random forests. Each one has its strengths and weaknesses. Logistic regression is easy to understand and quick to train, SVMs are great for finding clear decision boundaries, and random forests are good at handling complex patterns, but the best choice often depends on the size and nature of your data.

In this article, we’ll compare these three methods and see which one tends to work best for smaller datasets.

All three are quite reasonable algorithms to compare, though I’d want to add in gradient descent or XGBoost, as I’d expect it to perform better than random forest with small datasets.

Comments closed

Tracking Time Series Rates of Change in SQL Server

Published 2025-08-26 by Kevin Feasel

Rick Dobson wants a measure of variation:

This tip presents a brief introduction to Common Table Expressions (CTE), along with a few references for those seeking additional details on CTEs beyond those described and demonstrated here. We will examine CTEs that are defined by either one or two SELECT statements. Additionally, we will provide a demonstration of a recursive CTE. All the examples illustrate how to process time series datasets with CTEs.

Click through for the tip.

Comments closed

Passing Selections from Visuals to Translytical Task Flows

Published 2025-08-26 by Kevin Feasel

Jon Vöge sends along some data:

A common misconception about Translytical Task Flows is that the only way for you to parameterize and pass user inputs to the User Data Function, is through Slicers in Power BI.

That is not true at all.

In fact, one of the most powerful ways of integrating Task Flows into your Power BI reports, is by allowing user selections made in visualisations in your report, flow through to your task flow.

Read on to see how you can do this.

Comments closed

Microsoft Fabric Pipeline Copy Job Activity in Preview

Published 2025-08-26 by Kevin Feasel

Connie Xu makes an announcement:

We’re thrilled to announce that the Copy job Activity is now in Preview! 

This new orchestration activity brings the simplicity of the Copy job item directly into your Microsoft Fabric Data Factory pipelines, enabling you to manage data movement alongside transformations, notifications, and more; all in one place. 

Read on to learn more about it, including how it differs from the Copy activity and the Copy job item.

Comments closed

Week-over-Week Calculations in Power BI without DAX

Published 2025-08-26 by Kevin Feasel

Boniface Muchendu wants to avoid DAX for this calculation:

Power BI visual calculations provide a simplified way to perform time-based comparisons like week-over-week (WoW) analysis without writing complex DAX. Power BI’s time intelligence functions are powerful, but they can become challenging when working with custom calendars or unique business logic. If you’ve ever struggled with calculating WoW performance, this post will show you how to streamline the process using visual calculations.

Read on to see how. Though one thing to keep in mind is that it says week-over-week, but that assumes you have data for each week. I don’t believe this solution would cleanly handle gaps in data, so that would be something you’d want to do beforehand.

Comments closed

Join Operator Properties and Query Performance

Published 2025-08-26 by Kevin Feasel

Andy Brownsword takes a closer look at the big three join operators in SQL Server:

In the previous post I explained Join Operators in SQL Server. Whilst compiling that I dug a little deeper and came across a few interesting points I thought were worth sharing.

Let’s look at behaviour of the operators which may occur under specific conditions. Hopefully you find them as interesting as I did:

Click through for an interesting tidbit about nested loops joins, merge joins, and hash joins.

Comments closed

SQL Server 2025 RC0 Now Available

Published 2025-08-25 by Kevin Feasel

Priya Sathy has an announcement:

The first release candidate (RC0) of SQL Server 2025 is now available. As we move toward general availability, our focus shifts to delivering enhanced stability, performance, and product improvements based on your feedback.  

There are a few new features that Priya specifies in the post, so check those out. This also puts us on track for a GA delivery right around conference season.

Comments closed

Secrets Management in Kubernetes

Published 2025-08-25 by Kevin Feasel

Anshika Varshney takes us through some secrets management:

Secret management is a critical part of working with Kubernetes. When working with Kubernetes, you’ll often need to store and use sensitive information — like database passwords, API keys, and certificates. That’s where Kubernetes Secrets come in. But here’s the thing: while Kubernetes provides a way to store secrets, using them incorrectly can still expose your sensitive data.

This blog will walk you through:

  • What Kubernetes Secrets are
  • How they work
  • Common mistakes (the wrong way)
  • Best practices (the right way)
  • Real-life examples and configurations

Read on for the full article. I do want to iterate what I consider the most important of Anshika’s points: these things should never go into source control. The problem is, source control history is a lot more permanent than people customarily think about, so if you accidentally checked in a password in a config file somewhere, then got rid of the line with the password in it, a bad person with access to your source control can still look at history and find that password. The mechanism to rewrite Git history to remove a line’s existence is purposefully difficult, so a lot of developers don’t even know that it’s possible. But even then, if you’re hosting on someplace like GitHub and forks are enabled, rewriting history in your fork doesn’t rewrite it in somebody else’s fork, so again, a bad person with a fork from the right time frame could still see your exposed password.

The only fix is to cycle credentials if you ever do accidentally expose a password or key in source control files.

Comments closed

Row and Range Frames in Window Functions and Batch Mode

Published 2025-08-25 by Kevin Feasel

Erik Darling covers how your window frame (that is, ROWS or RANGE in the window function definition) can affect batch mode.

Erik looks at a classic performance difference between ROWS and RANGE, as well as what batch mode does to even the score. This is particularly nice because ROWS and RANGE both have their utility and focusing on one versus the other for performance differences can lead to awkward development practices to get around a window spool.

Erik also focuses primarily on batch mode on rowstore, so keep in mind the minimum requirements for it: 131,072 (or 2^17) rows in at least one table in the query, at least one operator that benefits from batch mode (which we’d cover in the window function), at least one input of the batch with 2^17 rows, and where the batch mode cost is lower than the row mode cost.

Comments closed

Microsoft Fabric Service Principal API Settings

Published 2025-08-25 by Kevin Feasel

Nicky van Vroenhoven has a public service announcement:

Microsoft Fabric is changing how service principal access to public APIs is controlled. The existing all-or-nothing tenant setting was split into two separate settings — giving us admins more granular control, but also introducing a change you might need to act on after August 1, 2025.

Click through to see how you might have been able to learn this, as well as the consequences of this change.

Comments closed