1+ USE [tpcxbb_1gb]
2+ GO
3+
4+ -- **************************************************************
5+ -- STEP 1 Create a table for storing the machine learning model
6+ -- **************************************************************
7+ DROP TABLE IF EXISTS [dbo].[models]
8+ GO
9+ CREATE TABLE [dbo].[models](
10+ [language] [varchar](30 ) NOT NULL ,
11+ [model_name] [varchar](30 ) NOT NULL ,
12+ [model] [varbinary](max ) NOT NULL ,
13+ [create_time] [datetime2](7 ) NULL DEFAULT (sysdatetime ()),
14+ [created_by] [nvarchar](500 ) NULL DEFAULT (suser_sname ()),
15+ PRIMARY KEY CLUSTERED
16+ (
17+ [language],
18+ [model_name]
19+ )
20+ )
21+ GO
22+
23+ -- *************************************************************************************************************
24+ -- STEP 2 Look at the dataset we will use in this sample
25+ -- Tag is a label indicating the sentiment of a review. These are actual values we will use to train the model
26+ -- For training purposes, we will use 90% percent of the data.
27+ -- For testing / scoring purposes, we will use 10% percent of the data.
28+ -- *************************************************************************************************************
29+ CREATE OR ALTER VIEW product_reviews_training_data
30+ AS
31+ SELECT TOP (CAST ( ( SELECT COUNT (* ) FROM product_reviews)* .9 AS INT ))
32+ CAST (pr_review_content AS NVARCHAR (4000 )) AS pr_review_content,
33+ CASE
34+ WHEN pr_review_rating < 3 THEN 1
35+ WHEN pr_review_rating = 3 THEN 2
36+ ELSE 3
37+ END AS tag
38+ FROM product_reviews;
39+ GO
40+
41+ CREATE OR ALTER VIEW product_reviews_test_data
42+ AS
43+ SELECT TOP (CAST ( ( SELECT COUNT (* ) FROM product_reviews)* .1 AS INT ))
44+ CAST (pr_review_content AS NVARCHAR (4000 )) AS pr_review_content,
45+ CASE
46+ WHEN pr_review_rating < 3 THEN 1
47+ WHEN pr_review_rating = 3 THEN 2
48+ ELSE 3
49+ END AS tag
50+ FROM product_reviews;
51+ GO
52+
53+ -- Look at the dataset we will use in this sample
54+ SELECT TOP (100 ) * FROM product_reviews_training_data;
55+ GO
56+
57+ -- ***************************************************************************************************
58+ -- STEP 3 Create a stored procedure for training a
59+ -- text classifier model for product review sentiment classification (Positive, Negative, Neutral)
60+ -- 1 = Negative, 2 = Neutral, 3 = Positive
61+ -- ***************************************************************************************************
62+ CREATE OR ALTER PROCEDURE [dbo].[create_text_classification_model]
63+ AS
64+ BEGIN
65+ DECLARE @model varbinary (max )
66+ , @train_script nvarchar (max );
67+
68+ -- The Python script we want to execute
69+ SET @train_script = N'
70+ ##Import necessary packages
71+ from microsoftml import rx_logistic_regression,featurize_text, n_gram
72+ import pickle
73+
74+ ## Defining the tag column as a categorical type
75+ training_data["tag"] = training_data["tag"].astype("category")
76+
77+ ## Create a machine learning model for multiclass text classification.
78+ ## We are using a text featurizer function to split the text in features of 2-word chunks
79+ model = rx_logistic_regression(formula = "tag ~ features", data = training_data, method = "multiClass", ml_transforms=[
80+ featurize_text(language="English",
81+ cols=dict(features="pr_review_content"),
82+ word_feature_extractor=n_gram(2, weighting="TfIdf"))])
83+
84+ ## Serialize the model so that we can store it in a table
85+ modelbin = pickle.dumps(model)
86+ '
87+
88+ EXECUTE sp_execute_external_script
89+ @language = N ' Python'
90+ , @script = @train_script
91+ , @input_data_1 = N ' SELECT * FROM product_reviews_training_data'
92+ , @input_data_1_name = N ' training_data'
93+ , @params = N ' @modelbin varbinary(max) OUTPUT'
94+ , @modelbin = @model OUTPUT ;
95+
96+ -- Save model to DB Table
97+ DELETE FROM dbo .models WHERE model_name = ' rx_logistic_regression'and language = ' Python'
98+ INSERT INTO dbo .models (language , model_name, model) VALUES (' Python'' rx_logistic_regression'
99+ END ;
100+ GO
101+
102+ -- ***************************************************************************************************
103+ -- STEP 4 Execute the stored procedure that creates and saves the machine learning model in a table
104+ -- ***************************************************************************************************
105+
106+ EXECUTE [dbo].[create_text_classification_model];
107+ -- Take a look at the model object saved in the model table
108+ SELECT * FROM dbo .models ;
109+ GO
110+
111+ -- ******************************************************************************************************************
112+ -- STEP 5 --Stored procedure that uses the model we just created to predict/classify the sentiment of product reviews
113+ -- ******************************************************************************************************************
114+ CREATE OR ALTER PROCEDURE [dbo].[predict_review_sentiment]
115+ AS
116+ BEGIN
117+ -- text classifier for online review sentiment classification (Positive, Negative, Neutral)
118+ DECLARE
119+ @model_bin varbinary (max )
120+ , @prediction_script nvarchar (max );
121+
122+ -- Select the model binary object from the model table
123+ SET @model_bin = (select model from dbo .models WHERE model_name = ' rx_logistic_regression'and language = ' Python'
124+
125+
126+ -- The Python script we want to execute
127+ SET @prediction_script = N'
128+ from microsoftml import rx_predict
129+ from revoscalepy import rx_data_step
130+ import pickle
131+
132+ ## The input data from the query in @input_data_1 is populated in test_data
133+ ## We are selecting 10% of the entire dataset for testing the model
134+
135+ ## Unserialize the model
136+ model = pickle.loads(model_bin)
137+
138+ ## Use the rx_logistic_regression model
139+ predictions = rx_predict(model = model, data = test_data, extra_vars_to_write = ["tag", "pr_review_content"], overwrite = True)
140+
141+ ## Converting to output data set
142+ result = rx_data_step(predictions)
143+ '
144+
145+ EXECUTE sp_execute_external_script
146+ @language = N ' Python'
147+ , @script = @prediction_script
148+ , @input_data_1 = N ' SELECT * FROM product_reviews_test_data'
149+ , @input_data_1_name = N ' test_data'
150+ , @output_data_1_name = N ' result'
151+ , @params = N ' @model_bin varbinary(max)'
152+ , @model_bin = @model_bin
153+ WITH RESULT SETS ((" Review" NVARCHAR (MAX )," Tag" FLOAT , " Predicted_Score_Negative" FLOAT , " Predicted_Score_Neutral" FLOAT , " Predicted_Score_Positive" FLOAT ));
154+ END
155+ GO
156+
157+
158+ -- ***************************************************************************************************
159+ -- STEP 6 Execute the multi class prediction using the model we trained earlier
160+ -- ***************************************************************************************************
161+ EXECUTE [dbo].[predict_review_sentiment]
162+ GO
163+
164+
165+
166+
167+
0 commit comments