I have a complex data model consisting of around hundred tables containing business data. Some tables are very wide, up to four hundred columns. Columns can have various data types - integers, decimals, text, dates etc. I'm looking for a way to identify relevant / important information stored in these tables.
I fully understand that business knowledge is essential to correctly process a data model. What I'm looking for are some strategies to pre-process tables and identify columns that should be taken to later stage where analysts will actually look into it. For example, I could use data profiling and statistics to find and exclude columns that don't have any data at all. Or maybe all records have the same value. This way I could potentially eliminate 30% of fields. However, I'm interested in exploring how AI and Machine Learning techniques could be used to identify important columns, hoping I could identify around 80% of relevant data. I'm aware, that relevant information will depend on the questions I want to ask. But even then, I hope I could narrow the columns to simplify the manual assesment taking place in the next stage.
Could anyone provide some guidance on how to use AI and Machine Learning to identify relevant columns in such wide tables? What strategies and techniques can be used to pre-process tables and identify columns that should be taken to the next stage?
Any help or guidance would be greatly appreciated. Thank you.
F.
The most common approach I've seen to evaluate the analytical utility of columns is the correlation method. This would tell you if there is a relationship (positive or negative) among specific column pairs. In my experience you'll be able to more easily build analysis outputs when columns are correlated - although, these analyses may not always be the most accurate.
However, before you even do that, like you indicate, you would probably need to narrow down your list of columns using much simpler methods. For example, you could surely eliminate a whole bunch of columns based on datatype and basic count statistics.
Less common analytic data types (ids, blobs, binary, etc) can probably be excluded first, followed by running simple COUNT(Distinct(ColName)), and Count(*) where ColName is null . This will help to eliminating UniqueIDs, Keys, and other similar data types. If all the rows are distinct, this would not be a good field for analysis. Same process for NULLs, if the percentage of nulls is greater than some threshold then you can eliminate those columns as well.
In order to automate it depending on your database, you could create a fairly simple stored procedure or function that loops through all the tables and columns and does a data type, count_distinct and a null percentage analysis on each field.
Once you've narrowed down list of columns, you can consider a .corr() function to run the analysis against all the remaining columns in something like a Python script.
If you wanted to keep everything in the database, Postgres also supports a corr() aggregate function, but you'll only be able to run this on 2 columns at a time, like this:
SELECT corr(column1,column2) FROM table;
so you'll need to build a procedure that evaluates multiple columns at once.
Thought about this tech challenges for some time. In general it’s AI solvable problem since there are easy features to extract such as unique values, clustering, distribution, etc.
And we want to bake this ability in https://columns.ai, obviously we haven’t gotten there yet, the first step we have done though is to collect all columns stats upon a data connection, identify columns that have similar range of unique values and generate a bunch of query templates for users to explore its dataset.
If interested, please take a look, as we keep advancing this part, it will become closer to an AI model to find relevant columns. Cheers!
Related
I'm investigating data warehouses. And I have an issue about star schemas.
It's in
Oracle® OLAP Application Developer's Guide
10g Release 1 (10.1)
3.2.1 Dimension Table: TIME_DIM
https://docs.oracle.com/cd/B13789_01/olap.101/b10333/global.htm#CHDCGABE
To represent the hierarchy MONTH -> QUARTER -> YEAR, we need some keys such as: YEAR_ID, QUARTER_ID. But there are some things that I do not understand:
1) Why do we need field YEAR_DSC & QUARTER_DSC? I think that we can look up these values from YEAR & QUARTER TABLE. And it breaks 2NF.
2) What is the normal form that a schema in data warehouse needs to satisfy? (1NF, 2NF, 3NF, or any.)
NFs (normal forms) don't matter for data warehouse base tables.
We normalize to reduce certain kinds of redundancy so that when we update a database we don't have to say the same thing in multiple places and so that we can't accidentally erroneously not say the same thing where it would need to be said in multiple places. That is not a problem in query results because we are not updating them. The same is true for a data warehouse's base tables. (Which are also just queries on its original database's base tables.)
Data warehouses are usually optimized for reading speed, and that usually means some denormalization compared to the original database to avoid recomputation at the expense of space. (Notice though that sometimes rereading something bigger can be slower than reading smaller parts and recomputing the big thing.) We probably don't want to drop normalized tables when moving to a data warehouse, because they answer simple queries and we don't want to slow down by recomputing them. Other than those tradeoffs, there's no reason not to denormalize. Some particular warehouse design methods might have their own rules about what parts should be denormalized what amounts.
(Whatever our original database design NF is chosen to be, we should always first normalize to 5NF then consciously denormalize. We don't need to normalize or know constraints to update or query a database.)
Read some textbook basics on why we normalize & why we use data warehouses.
I'm trying to design a Data Warehouse for a single store of commonly required data ranging from finance systems, project scheduling systems and a myriad of scientific systems. I.e. many different data marts.
I have been reading up on Data Warehousing and popular methods such as Star Schemas and Kimball methods etc but one question I cannot find answer to is:
Why is it better to design your DW Data Mart as a star schema rather than a single flat table?
Surely having no joins between facts and attributes/dimensions is faster and simpler than having lots of small joins to all the dimension tables? Disk space is not a problem, we'll just throw more disks at the database if necessary. Is the star schema slightly outdated these days or is it still data architect dogma?
Your question is very good: the Kimball mantra for dimensional modelling is to improve performance and to improve usability.
But I don't think it is outdated, or dogma- it is a reasonable, practical approach for many situations and platforms.
The way relational DBs store data means there's a balancing act to be struck between the numbers and types of tables, the routes in to the data for typical queries, easy maintainability and description of relationships between data, the numbers of joins, the way the joins are constructed, the indexability of columns, etc.
3NF (or further) is one end of the spectrum, suiting OLTP systems, and a single table is the other end of the spectrum. Dimensional models are in the middle and appropriate for reporting, at least when using certain technologies.
Performance isn't all about 'number of joins', although a star schema performs better for reporting workloads than a fully normalised database, in part because of a reduce number of joins. Dimensions are typically very wide. If you are including all those dimension fields in every row of every fact, you have very large rows indeed, and finding your way into those rows will perform very badly for typical queries.
Facts are numerous, so if you can make those tables compact, with the 'wordier' dimensions filterable, you hit a sweet spot of performance that a single table isn't going to match, unless heavily indexed.
And yes a single table for a fact is simpler in terms of numbers of tables but is it really easier to navigate? Dimensions and facts are easy concepts to understand, and what if you want to cross you queries across facts? You've got many different data marts but one of the benefits of having a data warehouse in the first place is that these aren't distinct- they're related and can be reported across. Conformed dimensions enable this.
If you combine your fact and dimensions into a single table, you'll either lose the visibility into dimension attributes that have never been used, or your measures will be thrown off by inclusion of a dummy event for the unused dimension attribute.
For example, a restaurant menu is a dimension and the purchased food is a fact. If you combined these into one table, how would you identify which food has never been ordered? For that matter, prior to your first order, how would you identify what food was available on the menu?
The dimension represents the possibilities, the fact represents the realization of the possibilities.
Combining facts and dimensions in the same table limits the scalability and the flexibility.
Suppose that one day the business decides to change a dimension description ( for example the product name ). Dimension tables aren't as deep as the fact tables and the update process or SCD management should be easier and less resource intensive.
I'm struggling to find a real world example on how to use google cloud dataflow combiners to run a common ETL tasl which aggregates records on multiple keys (e.g. Date, Location) and sums values over different measures (e.g. GrossValue, NetValue, Quantity). I can only find examples with a typical Key/Value (e.g. Day/Value) aggregation. Any hints on how this is done with the Python SDK would be appreciated.
I'm not 100% sure I understand your question. Do you have separate elements you are trying to join the data together for, in which case you may wish to use CoGroupByKey? Or does a single element have multiple fields?
Hope some of this info helps,
I would suggest looking at windowing, which will allow you to subdivide a PCollection according to the timestamps of its individual elements. If you want to see all the events for particular day this may be useful. Python examples of windowing. You may want to window across a days worth of data. This link is useful as well to understand how you can use GroupByKey in different ways,
Another option is to determine what date your elements belongs to, and use a group by key to key it with "[location][date][other]". You may need to do something like this if you want to join the data based on multiple fields.
See this GroupByKey example, but change the key to use your multiple fields concatenated.
Here is an example for reducing with a custom combiner. You can add logic here to do a custom aggregation for multiple different measurements.
I am designing a system for anomaly detection.
There are multiple approaches for building such system. I choose to implement one facet of such system by detection of features shared by the majority of samples. I acknowledge the possible insufficiencies of such method but for my specific use-case: (1) It suffices to know that a new sample contains (or lacks) features shared by the majority of past data to make a quick decision.(2) I'm interested in the insights such method will offer to the data.
So, here is the problem:
Consider a large data set with M data points, where each data point may include any number of {key:value} features. I choose to model a training dataset by grouping all the features observed in the data (the set of all unique keys) and setting it as the model's feature space. I define each sample by setting its values for existing keys and None for values in features it does not include.
Given this training data set I want to determine which features reoccur in the data; and for such reoccurring features, do they mostly share a single value.
My question:
A simple solution would be to count everything - for each of the N features calculate the distribution of values. However as M and N are potentially large, I wonder if there is a more compact way to represent the data or more sophisticated method to make claims about features' frequencies.
Am I reinventing an existing wheel? If there's an online approach for accomplishing such task it would be even better.
If I understand correctly your question,
you need to go over all the data anyway, so why not using hash?
Actually two hash tables:
Inner hash table for the distribution of feature values.
Outer hash table for feature existence.
In this way, the size of the inner hash table will indicate how is the feature common in your data, and the actual values will indicate how they differ one another. Another thing to notice is that you go over your data only once, and the time complexity for every operation (almost) on hash tables (if you allocate enough space from the beginning) is O(1).
Hope it helps
Is there a way to compare all information (aggregates, down to the detail level) between two OLAP cubes? For example, say I wanted to compare one cube created to work with sql server 2000 to that same cube, but migrated to run on sql server 2005/2008 - technically they should both return the same information for all dimension / measure combinations but I need a way to verify.
I am definitely NOT a developer, but I do have access to enterprise manager, and potentially SAS tools etc. and I know a bit of SQL but not much else. I know that you can compare two dimensional (i.e. tables) data sets with sql queries, and also with SAS - but I have never heard of a way to compare three dimensional cubes.
Am I out of luck on this one? The last thing that I want to have to do is view both cubes and compare all possible results side by side via excel or something, I hope that it can be automated somehow.
Comparing cubes means doing enough "slice-and-dice" queries to prove that you've queried all of the facts.
You can, simply, get a sum and count of the various fact and dimension tables. If those are the same, odds are good that any particular query will be the same between the two.
Without details on the dimensions and facts in question, it's hard to make a more specific recommendation.
However, consider that you can easily compute a set of subtotals for each dimension of the cube. If the dimensions are the same number of rows, the results will be the same number of rows. If the grand total is the same, then all that's left is row-by-row comparison of the subtotals.
If you do this once for each dimension, you should have some confidence that they're the same. Or, you'll find a difference that you can explore with more detailed queries.
The best approach is to compare the cube data by interchanging the rows and columns and verifying if all the counts and totals match properly.
For example, if you are having year-wise totals for a particular location, it would be a good approach to interchange the values between locations and the months and verifying whether they match properly.