• Olaf Nimz

    Sensor extended by dynamic local component – Kalman Filter with KFAS

    Linear models capture well the relations between sensor data if the system is in a steady state. Establishing relationships on the same row of a data set means no time dependency. Even systems which are switched on and off during working periods can be...
  • Olaf Nimz

    unpivot table in DAX as preparation for Horizon Plot in Power BI

    Horizon Plot requires a single column with measured values as main input. Typical data tables separate each Sensor in a separate column. To keep the first table and generate the appropriate format for Horizon Plot, a few lines of DAX reformat, aggregate...
  • Olaf Nimz

    Horizon Plot in Power BI

    Custom visuals for Power BI complement the bullet-proof standard visuals by interesting graphical capabilities. The Horizon plot is very useful for getting a quick impression of multivariate time series. The image below shows a panel with timebrush visual...
  • Olaf Nimz

    Calculate 1st derivative – local difference or slope – in PowerBI for differential equation

    On the way to data driven «model discovery for dynamical systems» by solving Ax = b with the b vector represents the time dependent change or dy/dt derivative of the system output (see Nathan Kutz on youtube for details). To visualize the dynamics of...
  • Olaf Nimz

    Outlier Detection in WorldWideImporters DWH

    As in excel with their „highlight exception“ function in Data Mining add-in one can check for extraordinary rows e.g. in a fact table as part of the star schema of the DWH for WWI just on a bigger scale. Building the clustering in SSAS dimensional...
  • Olaf Nimz

    can’t see tables due to timeout – why some transactions (b)lock exclusively

    Facing the annoying situation that management studio won’t show you the tables of a database but complain about timeout the usual suspects are:  transactions. But exclusive locks – blocking (protecting) the integrity of tables – are...
  • Olaf Nimz

    deep learning of time series

    The recent interview in O’Reilly’s Data Show Podcast covers in-memory streaming technologies like Apache Spark, Alluxio, and other open source technologies. More interestingly, deep learning is recommended for time series analysis. Particularly...
  • Olaf Nimz

    F1 Score – Formula One for choosing the most suitable Model

    F1 is a diagnostic tool with fine-tuned balance between ying and yang of precision and recall. The new episode of  Data Sceptic Podcast illustrates its utility in a plausible story. They tell us about the vivid analogy to design choices and the typical...
  • Olaf Nimz

    Basics of Maximum Likelihood explained

    The podcast replay of episode „How to Learn Statistical Regularities using Maximum Likelihood and Maximum A Posteriori Estimation“ from Machine Leaning 101 clarifies the basic concept how to learn probabilistic rules of a statistical environment...
  • Olaf Nimz

    Polybase cannot connect to HDInsight !

    All attempts to connect HDInsight via Polybase must fail as MS does not support its own Hadoop Cloud PAAS. So there is no LOCATION = hdfs://HeadNodeIP:Port that will work. Even if the MOC 20467C „Designing Self-Service Business Intelligence and...
  • Olaf Nimz

    import logins of active directory groups

    The issue of importing user logins from AD to fill a dimension of users for data security is twofold. First the paging limit of 1000 rows and second the hierarchy by implicit parent child relation of AD groups to their member login accounts which might...
  • Olaf Nimz

    what if – you want to setup in-Database Analytics in SQL Server 2016

    Just imagine you want to the in-database analytics feature of 2016 as a scoring engine. Meaning – setting up the additional functionality for advanced analytics on the server and on your  laptop the required tools to push your predictive modeling...
  • Olaf Nimz

    First impression of the R integration in SQL Server 2016

    Advanced Analytics Extension is one of the major features of SQL Server 2016 which is finally released with community technical preview CTP3 to the public. The expectation of in-database analytics require a seamless execution of R functionality on the...
  • Olaf Nimz

    Linking an Partition Schema to Heap Table

    Database projects in Visual Studio Shell or Data Tools (SSDT) provides additional functionality during refactoring of an existing DWH. Especially determining object dependencies between databases via searching the entire project incl. SSIS packages and...
  • Olaf Nimz

    fooled by testing for uniqueness – GUIDs are evil, if …

    Sometimes the identifier key provided in a source system is a kind of hash i.e. something like a GUID or Globally Unque IDentifier. Usually  – due to performance reasons – randomly generated keys are not preferable choice as a primary key...
  • Olaf Nimz

    flatten a table with dynamic sql pivot

    The task to rearrange a table with separate rows for each measurement to individual columns for every performance measure can be achieved in sql with the command pivot. The extension of SQL for Data Mining in SSAS cubes (DMX) provide a FLATTENED command...
Page 1 of 1 (16 items)