oracle data mining
Normalization is the process of scaling continuous values down to a specific range, often between zero and one. Normalization transforms each numerical value by subtracting a number, called the shift, and dividing the result by another number called the scale. The normalization techniques include:
- Min-Max Normalization : There is where the normalization is based on the using the minimum value for the shift and the (maximum-minimum) for the scale.
- Scale Normalization : This is where the normalization is based on zero being used for the shift and the value calculated using max[abs(max), abs(min)] being used for the scale
- Z-Score Normalization : This is where the normalization is based on using the mean value for the shift and the standard deviation for the scale.
When using Automatic Data Processing the normalization functions are used. But sometimes you may want to process the data is a more explicit manner. To do so you can use the various normalization function. To use these there is a three stage process. The first stage involves the creation of a table that will contain the normalization transformation data. The second stage applies the normalization procedures to your data source, defines the normalization required and inserts the required transformation data into the table create during the first stage. The third stage involves the defining of a view that applies the normalization transformations to your data source and displays the output via a database view. The following example illustrates how you can normalize the AGE and YRS_RESIDENCE attributes. The input data source will be the view that was created as the output of the previous transformation (MINING_DATA_V_2). This is passed on the original MINING_DATA_BUILD_V data set. The final output from this transformation step and all the other data transformation steps is MINING_DATA_READY_V.
BEGIN -- Clean-up : Drop the previously created tables BEGIN execute immediate 'drop table TRANSFORM_NORMALIZE'; EXCEPTION WHEN others THEN null; END; -- Stage 1 : Create the table for the transformations -- Perform normalization for: AGE and YRS_RESIDENCE dbms_data_mining_transform.CREATE_NORM_LIN ( norm_table_name => 'MINING_DATA_NORMALIZE'); -- Step 2 : Insert the normalization data into the table dbms_data_mining_transform.INSERT_NORM_LIN_MINMAX ( norm_table_name => 'MINING_DATA_NORMALIZE', data_table_name => 'MINING_DATA_V_2', exclude_list => DBMS_DATA_MINING_TRANSFORM.COLUMN_LIST ( 'affinity_card', 'bookkeeping_application', 'bulk_pack_diskettes', 'cust_id', 'flat_panel_monitor', 'home_theater_package', 'os_doc_set_kanji', 'printer_supplies', 'y_box_games')); -- Stage 3 : Create the view with the transformed data DBMS_DATA_MINING_TRANSFORM.XFORM_NORM_LIN ( norm_table_name => 'MINING_DATA_NORMALIZE', data_table_name => 'MINING_DATA_V_2', xform_view_name => 'MINING_DATA_READY_V'); END; /
The above example performs normalization based on the Minimum-Maximum values of the variables/columns. The other normalization functions are:
|INSERT_NORM_LIN_SCALE||Inserts linear scale normalization definitions in a transformation definition table.|
|INSERT_NORM_LIN_ZSCORE||Inserts linear zscore normalization definitions in a transformation definition table.|
Everyday someone talks about the the processing power needed for Machine Learning, and the vast computing needed for these tasks. It has become evident that most of these people have never created a machine learning model. Never. But like to make up stuff and try to make themselves look like an expert, or as I and others like to call them a “fake expert”.
When you question these “fake experts” about this topic, they huff and puff about lots of things and never answer the question or try to claim it is so difficult, you simply don’t understand.
Having worked in the area of machine learning for a very very long time, I’ve never really had performance issues with creating models. Yes most of the time I’ve been able to use my laptop. Yes my laptop to build models large models. In a couple of these my laptop couldn’t cope and I moved onto a server.
But over the past few years we keep hearing about using cloud services for machine learning. If you are doing machine learning you need to computing capabilities that are available with cloud services.
So, the results below show the results of building machine learning models, using different algorithms, with different sizes of data sets.
For this test, I used a basic cloud service. Well maybe it isn’t basic, but for others they will consider it very basic with very little compute involved.
I used an Oracle Cloud DBaaS for this experiment. I selected an Oracle 18c Extreme edition cloud service. This comes with the in-database machine learning option. This comes with 1 OCPUs, 7.5G Memory and 170GB storage. This is the basic configuration.
Next I created data sets with different sizes. These were based on one particular data set, as this ensures that as the data set size increases, the same kind of data and processing required remained consistent, instead of using completely different data sets.
The data set consisted of the following number of records, 72K, 660K, 210K, 2M, 10M and 50M.
I then created machine learning models using Decisions Tree, Naive Bayes, Support Vector Machine, Generaliszd Linear Models (GLM) and Neural Networks. Yes it was a typical classification problem.
The following table below shows the length of time in seconds to build the models. All data preparations etc was done prior to this.
Note: It should be noted that Automatic Data Preparation was turned on for these algorithms. This performed additional algorithm specific data preparation for each model. That means the times given in the following tables is for some data preparation time and for building the models.
Converting the above table into minutes.
As your company evolves with their data mining projects, the number of models produced and in use in production will increase dramatically.
Care needs to be taken when it comes to managing these. This includes using meaningful names, adding descriptions of what the model is about or for, and being able to track their usage, etc.
I will look at tracking the usage of the models in another blog post, but the following gives examples of how to rename Oracle Data Mining models and how to add comments or descriptions to these models. This is particularly useful because our data analytics teams have a constant turn over or it has been many months since you last worked on a model and you want a quick idea of what purpose of the model was for.
If you have been using the Oracle Data Mining tool (part of SQL Developer) will will see your model being created with some sort of sequencing numbers. For example for a Support Vector Machine (SVM) model you might see it labelled for classification:
While you are working on this project you will know and understand what it was about and why it is being used. But afterward you may forget as you will be dealing with many hundreds of models. Yes you could check your documentation for the purpose of this model but that can take some time.
What if you could run a SQL query to find out?
But first we need to rename the model.
Next we will want to add a longer description of what the model is about. We can do this by adding a comment to the model.
COMMENT ON MINING MODEL high_value_churn_clas_svm IS 'Classification Model to Predict High Value Customers most likely to Churn';
We can now see these updated details when we query the Oracle Data Mining models in a user schema.
SELECT model_name, mining_function, algorithm, comments FROM user_mining_models;
These are two very useful commands.
When working with the Clustering algorithms, and particularly k-Means, in the Oracle Data Miner tool there is no way of seeing how compact or dispersed the data is within a cluster.
There are a number of measures typically used in various tools and algorithms, but with Oracle Data Miner we are not presented with any of this information.
But if we flip from using the Oracle Data Miner tool to using SQL we can get to see some more details of the clusters produced by the k-Means algorithm along with some additional and useful information.
As I said there are a number of different measures used to evaluate clusters. The one that Oracle uses is called Dispersion. Now there are a few different definitions of what this could be and I haven’t been able to locate what is Oracle’s own definition of it in any of the documentation.
We can use the Dispersion value as a measure of how compact or how spread out the data is within a cluster. The Dispersion value is a number greater than 0. The lower the value of the more compact the cluster is i.e. the data points are close the the centroid of the cluster. The larger the value the more disperse or spread out the data points are.
The DBMS_DATA_MINING PL/SQL package comes with a function called GET_MODEL_DETAILS_KM. This function returns a record of the form DM_CLUSTERS.
(id NUMBER, cluster_id VARCHAR2(4000), record_count NUMBER, parent NUMBER, tree_level NUMBER, dispersion NUMBER, split_predicate DM_PREDICATES, child DM_CHILDREN, centroid DM_CENTROIDS, histogram DM_HISTOGRAMS, rule DM_RULE)
We can not use the following query to get the Dispersion value for each of the clusters from an ODM cluster model.
SELECT cluster_id, record_count, parent, tree_level, dispersion FROM table(dbms_data_mining.get_model_details_km('CLUS_KM_3_2'));
If you are a user of the Oracle Data Miner tool (the workflow data mining tool that is part of SQL Developer), then you will have noticed that for many of the algorithms you can specify a Case Id attribute along with, say, the target attribute.
The idea is that you have one attribute that is a unique identifier for each case record. This may or may not be the case in your data model and you may have a multiple attribute primary key or case record identifier.
But what is the Case Id field used for in Oracle Data Miner?
Based on the documentation this field does not need to have a value. But it is recommended that you do identify an attribute for the Case Id, as this will allow for reproducible results. What this means is that if we run our workflow today and again in a few days time, on the exact same data, we should get the same results. So the Case Id allows this to happen. But how? Well it looks like the attribute used or specified for the Case Id is used as part of the Hashing algorithm to partition the data into a train and test data set, for classification problems.
So if you don’t have a single attribute case identifier in your data set, then you need to create one. There are a few options open to you to do this.
- Create one: write some code that will generate a unique identifier for each of your case records based on some defined rule.
- Use a sequence: and update the records to use this sequence.
- Use ROWID: use the unique row identifier value. You can write some code to populate this value into an attribute. Or create a view on the table containing the case records and add a new attribute that will use the ROWID. But if you move the data, then the next time you use the view then you will be getting different ROWIDs and that in turn will mean we may have different case records going into our test and training data sets. So our workflows will generate different results. Not what we want.
- Use ROWNUM: This is kind of like using the ROWID. Again we can have a view that will select ROWNUM for each record. Again we may have the same issues but if we have our data ordered in a way that ensures we get the records returned in the same order then this approach is OK to use.
- Use Identity Column: In Oracle 12c we have a new feature called Identify Column. This kind of acts like a sequence but we can defined an attribute in a table to be an Identity Column, and as records are inserted into the the data (in our scenario our case table) then this column will automatically generate a unique number for our data. Again if we need to repopulate the case table, you will need to drop and recreate the table to get the Identity Column to reset, otherwise the newly inserted records will start with the next number of the Identity Column
Here is an example of using the Identity Column in a case table.
CREATE TABLE case_table ( id_column NUMBER GENERATED ALWAYS AS IDENTITY, affinity_card NUMBER, age NUMBER, cust_gender VARCHAR2(5), country_name VARCHAR2(20) ... );
You can now use this Identity Column as the Case Id in your Oracle Data Miner workflows.
In this 5th blog post in my series on using the capabilities of Oracle Text, Oracle R Enterprise and Oracle Data Mining to process documents and text, I will have a look at some of the machine learning features of Oracle Text.
Oracle Text comes with a number of machine learning algorithms. These can be divided into two types. The first is called ‘Supervised Learning’ where we have two machine learning algorithms for classification type of problem. The second type is called ‘Unsupervised Learning’ where we have the ability to use clustering machine learning algorithms to look for patterns in our text documents and to find similarities between documents based on their contents.
It is this second type of document clustering that I will work through in this blog post.
When using clustering with text documents, the machine learning algorithm will look for patterns that are common between the documents. These patterns will include the words used, the frequency of the words, the position or ordering of these words, the co-occurance of words, etc. Yes this is a large an complex task and that is why we need a machine learning algorithm to help us.
With Oracle Text we only have one clustering machine learning algorithm available to use. When we move onto using the Oracle Advanced Analytics Option (Oracle Data Mining and Oracle R Enterprise) we more algorithms available to us.
With Oracle Text the clustering algorithm is called k-Means. In a way the actual algorithm is unimportant as it is the only one available to us when using Oracle Text. To use this algorithm we have the CTX_CLS.CLUSTERING procedure. This procedure takes the documents we want to compare and will then identify the clusters (using hierarchical clustering) and will then tells us, for each document, what clusters the documents belong to and they probability value. With clustering a document (or a record) can belong to many clusters. Typically in the text books we see clusters that are very distinct and are clearly separated from each other. When you work on real data this is never the case. We will have many over lapping clusters and a data point/record can belong to one or more clusters. This is why we need the probability vale. We can use this to determine what cluster our record belongs to most and what other clusters it is associated with.
Using the example documents that I have been using during this series of blog posts we can use the CTX_CLS.CLUSTERING algorithm to cluster and identify similarities in these documents.
We need to setup the parameters that will be used by the CTX_CLS.CLUSTERING procedure. Tell it to use the k-Means algorithm and then the number of clusters to generate. As with all Oracle Text procedures or algorithms there are a number of settings you can configure or you can just accept the default values.
exec ctx_ddl.drop_preference('Cluster_My_Documents'); exec ctx_ddl.create_preference('Cluster_My_Documents','KMEAN_CLUSTERING'); exec ctx_ddl.set_attribute('Cluster_My_Documents','CLUSTER_NUM','3');
The code above is an example of the basics of what you need to setup for clustering. Other attribute or cluster parameter setting available to you include,
MAX_DOCTERMS, MAX_FEATURES, THEME_ON, TOKEN_ON, STEM_ON, MEMORY_SIZE and SECTION_WEIGHT.
Now we can run the CTX_CLS.CLUSTERING procedure on our documents. This procedure has the following parameters.
– The Oracle Text Index Name
– Document Id Column Name
– Document Assignment (cluster assignment) Table Name. This table will be created if it doesn’t already exist
– Cluster Description Table Name. This table will be created if it doesn’t already exist.
– Name of the Oracle Text Preference (list)
exec ctx_cls.clustering( 'MY_DOCUMENTS_OT_IDX', 'DOC_PK', 'OT_CLUSTER_RESULTS', 'DOC_CLUSTER_DETAILS', 'Cluster_My_Documents');
When the procedure has completed we can now examine the OT_CLUSTER_RESULTS and the DOC_CLUSTER_DETAILS tables. The first of these (OT_CLUSTER_RESULTS) allows us to see what documents have been clustered together. The following is what was produced for my documents.
SELECT d.doc_pk, d.doc_title, r.clusterid, r.score FROM my_documents d, ot_cluster_results r WHERE d.doc_pk = r.docid;
We can see that two of the documents have been grouped into the same cluster (ClusterId=2). If you have a look back at what these documents are about then you can see that yes these are very similar. For the other two documents we can see that they have been clustered into separate clusters (ClusterId=4 & 5). The clustering algorithms have said that they are different types of documents. Again when you examine these documents you will see that they are talking about different topics. So the clustering process worked !
You can also explore the various features of the clusters by looking that he DOC_CLUSTER_DETAILS table. Although the details in this table are not overly useful but it will give you some insight into what clusters the k-Means algorithm has produced.
Hopefully I’ve shown you how easy it is to setup and use the clustering feature of Oracle Text.
WARNING: Before using the Clustering or Classification with Oracle Text, you need to check with your local Oracle Sales representative about if there is licence implication. There seems to be some mentions the the algorithms used are those that come with Oracle Data Mining. Oracle Data Mining is a licence cost option for the database. So make sure you check before you go using these features.
A project that I’ve been working on for a while now involves the use of Oracle Text, Oracle R Enterprise and Oracle Data Mining. Oracle Text comes with your Oracle Database licence. Oracle R Enterprise and Oracle Data Mining are part of the Oracle Advanced Analytics (extra cost) option.
What I will be doing over the course of 4 or maybe 5 blog posts is how these products can work together to help you gain a grater insight into your data, and part of your data being large text items like free format text, documents (in various forms e.g. html, xml, pdf, ms word), etc.
Unfortunately I cannot show you examples from the actual project I’ve been working on (and still am, from time to time). But what I can do is to show you how products and components can work together.
In this blog post I will just do some data setup. As with all project scenarios there can be many ways of performing the same tasks. Some might be better than others. But what I will be showing you is for demonstration purposes.
The scenario: The scenario for this blog post is that I want to extract text from some webpages and store them in a table in my schema. I then want to use Oracle Text to search the text from these webpages.
Schema setup: We need to create a table that will store the text from the webpages. We also want to create an Oracle Text index so that this text is searchable.
drop sequence my_doc_seq; create sequence my_doc_seq; drop table my_documents; create table my_documents ( doc_pk number(10) primary key, doc_title varchar2(100), doc_extracted date, data_source varchar2(200), doc_text clob); create index my_documents_ot_idx on my_documents(doc_text) indextype is CTXSYS.CONTEXT;
In the table we have a number of descriptive attributes and then a club for storing the website text. We will only be storing the website text and not the html document (More on that later). In order to make the website text searchable in the DOC_TEXT attribute we need to create an Oracle Text index of type CONTEXT.
There are a few challenges with using this type of index. For example when you insert a new record or update the DOC_TEXT attribute, the new values/text will not be reflected instantly, just like we are use to with traditional indexes. Instead you have to decide when you want to index to be updated. For example, if you would like the index to be updated after each commit then you can create the index using the following.
create index my_documents_ot_idx on my_documents(doc_text) indextype is CTXSYS.CONTEXT parameters ('sync (on commit)');
Depending on the number of documents you have being committed to the DB, this might not be for you. You need to find the balance. Alternatively you could schedule the index to be updated by passing an interval to the ‘sync’ in the above command. Alternatively you might want to use DBMS_JOB to schedule the update.
To manually sync (or via DBMS_JOB) the index, assuming we used the first ‘create index’ statement, we would need to run the following.
This function just adds the new documents to the index. This can, over time, lead to some fragmentation of the index, and will require it to the re-organised on a semi-regular basis. Perhaps you can schedule this to happen every night, or once a week, or whatever makes sense to you.
BEGIN CTX_DDL.OPTIMIZE_INDEX('my_documents_ot_idx','FULL'); END;
(I could talk a lot more about setting up some basics of Oracle Text, the indexes, etc. But I’ll leave that for another day or you can read some of the many blog posts that already exist on the topic.)
Extracting text from a webpage using R: Some time ago I wrote a blog post on using some of the text mining features and packages in R to produce a word cloud based on some of the Oracle Advanced Analytics webpages.
I’m going to use the same webpages and some of the same code/functions/packages here.
The first task you need to do is to get your hands on the ‘htmlToText function. You can download the htmlToText function on github. This function requires the ‘Curl’ and ‘XML’ R packages. So you may need to install these.
I also use the str_replace_all function (“stringer’ R package) to remove some of the html that remains, to remove some special quotes and to replace and occurrences of ‘&’ with ‘and’.
# Load the function and required R packages
data1 <- str_replace_all(htmlToText("http://www.oracle.com/technetwork/database/options/advanced-analytics/overview/index.html"), "[\r\n\t\"\'\u201C\u201D]" , "") data1 <- str_replace_all(data1, "&", "and") data2 <- str_replace_all(str_replace_all(htmlToText("http://www.oracle.com/technetwork/database/options/advanced-analytics/odm/index.html"), "[\r\n\t\"\'\u201C\u201D]" , ""), "&", "and") data2 <- str_replace_all(data2, "&", "and") data3 <- str_replace_all(str_replace_all(htmlToText("http://www.oracle.com/technetwork/database/database-technologies/r/r-technologies/overview/index.html"), "[\r\n\t\"\'\u201C\u201D]" , ""), "&", "and") data3 <- str_replace_all(data3, "&", "and") data4 <- str_replace_all(str_replace_all(htmlToText("http://www.oracle.com/technetwork/database/database-technologies/r/r-enterprise/overview/index.html"), "[\r\n\t\"\'\u201C\u201D]" , ""), "&", "and") data4 <- str_replace_all(data4, "&", "and")
We now have the text extracted and cleaned up.
Create a data frame to contain all our data: Now that we have the text extracted, we can prepare the other data items we need to insert the data into our table (‘my_documents’). The first stept is to construct a data frame to contain all the data.
data_source = c("http://www.oracle.com/technetwork/database/options/advanced-analytics/overview/index.html", "http://www.oracle.com/technetwork/database/options/advanced-analytics/odm/index.html", "http://www.oracle.com/technetwork/database/database-technologies/r/r-technologies/overview/index.html", "http://www.oracle.com/technetwork/database/database-technologies/r/r-enterprise/overview/index.html") doc_title = c("OAA_OVERVIEW", "OAA_ODM", "R_TECHNOLOGIES", "OAA_ORE") doc_extracted = Sys.Date() data_text <- c(data1, data2, data3, data4) my_docs <- data.frame(doc_title, doc_extracted, data_source, data_text)
Insert the data into our database table: With the data in our data fram (my_docs) we can now use this data to insert into our database table. There are a number of ways of doing this in R. What I’m going to show you here is how to do it using Oracle R Enterprise (ORE). The thing with ORE is that there is no explicit functionality for inserting and updating records in a database table. What you need to do is to construct, in my case, the insert statement and then use ore.exec to execute this statement in the database.