The BOOLRULE procedure is a powerful tool for discovering 'IF... THEN...' rule relationships in datasets. In this example, we start by creating a table of customer reviews (`mycas.reviews`) including a binary sentiment ('positive'). Then, `PROC TEXTMINE` is employed to convert raw text into a structured representation (terms-documents matrix `mycas.reviews_bow` and terms table `mycas.reviews_terms`). Finally, `PROC BOOLRULE` is applied to generate predictive rules for sentiment based on the presence of certain terms or combinations of terms. Analyzing rule metrics (F1, Precision, Recall) helps evaluate the quality of these rules.
Data Analysis
Type : CREATION_INTERNE
Examples use internally generated data via DATALINES blocks or temporary CAS tables.
1 Code Block
PROC BOOLRULE Data
Explanation :
This example illustrates the minimal use of `PROC BOOLRULE` with a binary target. It follows the steps of data creation, text preprocessing with `PROC TEXTMINE`, then applies `PROC BOOLRULE` to extract rules. The `MINSUPPORTS`, `MPOS`, and `GPOS` options are set for basic control over rule generation.
This example illustrates the minimal use of `PROC BOOLRULE` with a binary target. It follows the steps of data creation, text preprocessing with `PROC TEXTMINE`, then applies `PROC BOOLRULE` to extract rules. The `MINSUPPORTS`, `MPOS`, and `GPOS` options are set for basic control over rule generation.
Copied!
| 1 | /* Établir une session CAS et créer la bibliothèque 'mycas' */ |
| 2 | cas; |
| 3 | LIBNAME mycas cas; |
| 4 | |
| 5 | /* Création de la table de données d'avis */ |
| 6 | DATA mycas.reviews; |
| 7 | INFILE DATALINES delimiter='|' missover; |
| 8 | LENGTH text $300 category $20; |
| 9 | INPUT text$ positive category$ did; |
| 10 | DATALINES; |
| 11 | This is the greatest phone ever! love it!|1|electronics|1 |
| 12 | The phone's battery life is too short and screen resolution is low.|0|electronics|2 |
| 13 | The screen resolution is low, but I love this tv.|1|electronics|3 |
| 14 | The movie itself is great and I like it, although the resolution is low.|1|movies|4 |
| 15 | The movie's story is boring and the acting is poor.|0|movies|5 |
| 16 | I watched this movie on tv, it's not good on a small screen. |0|movies|6 |
| 17 | watched the movie first and loved it, the book is even better!|1|books |7 |
| 18 | I like the story in this book, they should put it on screen.|1|books|8 |
| 19 | I love the author, but this book is a waste of time, don't buy it.|0|books|9 |
| 20 | ; |
| 21 | RUN; |
| 22 | |
| 23 | /* Prétraitement du texte avec PROC TEXTMINE */ |
| 24 | PROC TEXTMINE DATA=mycas.reviews; |
| 25 | doc_id |
| 26 | did; |
| 27 | var |
| 28 | text; |
| 29 | parse |
| 30 | nonoungroups |
| 31 | notagging |
| 32 | entities = none |
| 33 | outparent = mycas.reviews_bow |
| 34 | outterms = mycas.reviews_terms |
| 35 | reducef = 1; |
| 36 | RUN; |
| 37 | |
| 38 | /* Extraction des règles avec PROC BOOLRULE pour cible binaire 'positive' */ |
| 39 | PROC BOOLRULE |
| 40 | DATA = mycas.reviews_bow |
| 41 | docid = _document_ |
| 42 | termid = _termnum_ |
| 43 | docinfo = mycas.reviews |
| 44 | terminfo = mycas.reviews_terms |
| 45 | minsupports = 1 |
| 46 | mpos = 1 |
| 47 | gpos = 1; |
| 48 | docinfo |
| 49 | id = did |
| 50 | targets = (positive); |
| 51 | terminfo |
| 52 | id = key |
| 53 | label = term; |
| 54 | OUTPUT |
| 55 | ruleterms = mycas.ruleterms_basic |
| 56 | rules = mycas.rules_basic; |
| 57 | RUN; |
| 58 | |
| 59 | /* Affichage des règles générées */ |
| 60 | DATA rules_basic; |
| 61 | SET mycas.rules_basic; |
| 62 | RUN; |
| 63 | |
| 64 | PROC PRINT DATA=rules_basic; |
| 65 | title 'Exemple Basique: Règles pour sentiment positif'; |
| 66 | var target ruleid rule F1 precision recall; |
| 67 | RUN; |
| 68 | |
| 69 | /* Nettoyage */ |
| 70 | PROC CASUTIL; |
| 71 | droptable casdata='reviews' incaslib='mycas'; |
| 72 | droptable casdata='reviews_bow' incaslib='mycas'; |
| 73 | droptable casdata='reviews_terms' incaslib='mycas'; |
| 74 | droptable casdata='ruleterms_basic' incaslib='mycas'; |
| 75 | droptable casdata='rules_basic' incaslib='mycas'; |
| 76 | QUIT; |
| 77 |
2 Code Block
PROC BOOLRULE Data
Explanation :
This example extends the use of `PROC BOOLRULE` by adding common options to refine rule generation. `NOCUTOFF` prevents truncation of candidate rules, `MINRULELEN` and `MAXRULELEN` control rule complexity, and `NTHREADS` enables parallelism to improve performance.
This example extends the use of `PROC BOOLRULE` by adding common options to refine rule generation. `NOCUTOFF` prevents truncation of candidate rules, `MINRULELEN` and `MAXRULELEN` control rule complexity, and `NTHREADS` enables parallelism to improve performance.
Copied!
| 1 | /* Établir une session CAS et créer la bibliothèque 'mycas' */ |
| 2 | cas; |
| 3 | LIBNAME mycas cas; |
| 4 | |
| 5 | /* Création de la table de données d'avis */ |
| 6 | DATA mycas.reviews; |
| 7 | INFILE DATALINES delimiter='|' missover; |
| 8 | LENGTH text $300 category $20; |
| 9 | INPUT text$ positive category$ did; |
| 10 | DATALINES; |
| 11 | This is the greatest phone ever! love it!|1|electronics|1 |
| 12 | The phone's battery life is too short and screen resolution is low.|0|electronics|2 |
| 13 | The screen resolution is low, but I love this tv.|1|electronics|3 |
| 14 | The movie itself is great and I like it, although the resolution is low.|1|movies|4 |
| 15 | The movie's story is boring and the acting is poor.|0|movies|5 |
| 16 | I watched this movie on tv, it's not good on a small screen. |0|movies|6 |
| 17 | watched the movie first and loved it, the book is even better!|1|books |7 |
| 18 | I like the story in this book, they should put it on screen.|1|books|8 |
| 19 | I love the author, but this book is a waste of time, don't buy it.|0|books|9 |
| 20 | ; |
| 21 | RUN; |
| 22 | |
| 23 | /* Prétraitement du texte avec PROC TEXTMINE */ |
| 24 | PROC TEXTMINE DATA=mycas.reviews; |
| 25 | doc_id |
| 26 | did; |
| 27 | var |
| 28 | text; |
| 29 | parse |
| 30 | nonoungroups |
| 31 | notagging |
| 32 | entities = none |
| 33 | outparent = mycas.reviews_bow |
| 34 | outterms = mycas.reviews_terms |
| 35 | reducef = 1; |
| 36 | RUN; |
| 37 | |
| 38 | /* Extraction des règles avec des options supplémentaires */ |
| 39 | PROC BOOLRULE |
| 40 | DATA = mycas.reviews_bow |
| 41 | docid = _document_ |
| 42 | termid = _termnum_ |
| 43 | docinfo = mycas.reviews |
| 44 | terminfo = mycas.reviews_terms |
| 45 | minsupports = 1 |
| 46 | mpos = 1 |
| 47 | gpos = 1 |
| 48 | nocutoff /* Ne pas couper les règles candidates */ |
| 49 | minrulelen = 1 /* Longueur minimale de la règle */ |
| 50 | maxrulelen = 3 /* Longueur maximale de la règle */ |
| 51 | nthreads = 4; /* Utiliser 4 threads pour le traitement */ |
| 52 | docinfo |
| 53 | id = did |
| 54 | targets = (positive); |
| 55 | terminfo |
| 56 | id = key |
| 57 | label = term; |
| 58 | OUTPUT |
| 59 | ruleterms = mycas.ruleterms_intermediaire |
| 60 | rules = mycas.rules_intermediaire; |
| 61 | RUN; |
| 62 | |
| 63 | /* Affichage des règles générées */ |
| 64 | DATA rules_intermediaire; |
| 65 | SET mycas.rules_intermediaire; |
| 66 | RUN; |
| 67 | |
| 68 | PROC PRINT DATA=rules_intermediaire; |
| 69 | title 'Exemple Intermédiaire: Règles avec options courantes'; |
| 70 | var target ruleid rule F1 precision recall; |
| 71 | RUN; |
| 72 | |
| 73 | /* Nettoyage */ |
| 74 | PROC CASUTIL; |
| 75 | droptable casdata='reviews' incaslib='mycas'; |
| 76 | droptable casdata='reviews_bow' incaslib='mycas'; |
| 77 | droptable casdata='reviews_terms' incaslib='mycas'; |
| 78 | droptable casdata='ruleterms_intermediaire' incaslib='mycas'; |
| 79 | droptable casdata='rules_intermediaire' incaslib='mycas'; |
| 80 | QUIT; |
| 81 |
3 Code Block
PROC BOOLRULE Data
Explanation :
This example demonstrates a more advanced approach by filtering review data for a specific category ('electronics') before applying rule extraction. This allows for obtaining rules more relevant to a subset of the data, showing how `PROC BOOLRULE` can be integrated into a more complex data analysis workflow.
This example demonstrates a more advanced approach by filtering review data for a specific category ('electronics') before applying rule extraction. This allows for obtaining rules more relevant to a subset of the data, showing how `PROC BOOLRULE` can be integrated into a more complex data analysis workflow.
Copied!
| 1 | /* Établir une session CAS et créer la bibliothèque 'mycas' */ |
| 2 | cas; |
| 3 | LIBNAME mycas cas; |
| 4 | |
| 5 | /* Création de la table de données d'avis */ |
| 6 | DATA mycas.reviews; |
| 7 | INFILE DATALINES delimiter='|' missover; |
| 8 | LENGTH text $300 category $20; |
| 9 | INPUT text$ positive category$ did; |
| 10 | DATALINES; |
| 11 | This is the greatest phone ever! love it!|1|electronics|1 |
| 12 | The phone's battery life is too short and screen resolution is low.|0|electronics|2 |
| 13 | The screen resolution is low, but I love this tv.|1|electronics|3 |
| 14 | The movie itself is great and I like it, although the resolution is low.|1|movies|4 |
| 15 | The movie's story is boring and the acting is poor.|0|movies|5 |
| 16 | I watched this movie on tv, it's not good on a small screen. |0|movies|6 |
| 17 | watched the movie first and loved it, the book is even better!|1|books |7 |
| 18 | I like the story in this book, they should put it on screen.|1|books|8 |
| 19 | I love the author, but this book is a waste of time, don't buy it.|0|books|9 |
| 20 | ; |
| 21 | RUN; |
| 22 | |
| 23 | /* Filtrage des avis pour une catégorie spécifique (par exemple, 'electronics') */ |
| 24 | DATA mycas.reviews_filtered; |
| 25 | SET mycas.reviews; |
| 26 | where category = 'electronics'; |
| 27 | RUN; |
| 28 | |
| 29 | /* Prétraitement du texte avec PROC TEXTMINE sur les données filtrées */ |
| 30 | PROC TEXTMINE DATA=mycas.reviews_filtered; |
| 31 | doc_id |
| 32 | did; |
| 33 | var |
| 34 | text; |
| 35 | parse |
| 36 | nonoungroups |
| 37 | notagging |
| 38 | entities = none |
| 39 | outparent = mycas.reviews_bow_filtered |
| 40 | outterms = mycas.reviews_terms_filtered |
| 41 | reducef = 1; |
| 42 | RUN; |
| 43 | |
| 44 | /* Extraction des règles avec PROC BOOLRULE sur les données filtrées */ |
| 45 | PROC BOOLRULE |
| 46 | DATA = mycas.reviews_bow_filtered |
| 47 | docid = _document_ |
| 48 | termid = _termnum_ |
| 49 | docinfo = mycas.reviews_filtered |
| 50 | terminfo = mycas.reviews_terms_filtered |
| 51 | minsupports = 1 |
| 52 | mpos = 1 |
| 53 | gpos = 1; |
| 54 | docinfo |
| 55 | id = did |
| 56 | targets = (positive); |
| 57 | terminfo |
| 58 | id = key |
| 59 | label = term; |
| 60 | OUTPUT |
| 61 | ruleterms = mycas.ruleterms_avance |
| 62 | rules = mycas.rules_avance; |
| 63 | RUN; |
| 64 | |
| 65 | /* Affichage des règles générées */ |
| 66 | DATA rules_avance; |
| 67 | SET mycas.rules_avance; |
| 68 | RUN; |
| 69 | |
| 70 | PROC PRINT DATA=rules_avance; |
| 71 | title 'Exemple Avancé: Règles pour les avis électroniques'; |
| 72 | var target ruleid rule F1 precision recall; |
| 73 | RUN; |
| 74 | |
| 75 | /* Nettoyage */ |
| 76 | PROC CASUTIL; |
| 77 | droptable casdata='reviews' incaslib='mycas'; |
| 78 | droptable casdata='reviews_filtered' incaslib='mycas'; |
| 79 | droptable casdata='reviews_bow_filtered' incaslib='mycas'; |
| 80 | droptable casdata='reviews_terms_filtered' incaslib='mycas'; |
| 81 | droptable casdata='ruleterms_avance' incaslib='mycas'; |
| 82 | droptable casdata='rules_avance' incaslib='mycas'; |
| 83 | QUIT; |
| 84 |
4 Code Block
PROC BOOLRULE Data
Explanation :
This example highlights explicit CAS session management and parameter adjustment for performance and rule filtering. Support thresholds (`MINSUPPORTS`, `MPOS`, `GPOS`) are set proportionally, `MAXRULELEN` is used to limit rule complexity, and `NTHREADS` is increased to optimize execution in the distributed CAS environment. The input table is slightly enlarged to better simulate a real-world use case with large data volumes.
This example highlights explicit CAS session management and parameter adjustment for performance and rule filtering. Support thresholds (`MINSUPPORTS`, `MPOS`, `GPOS`) are set proportionally, `MAXRULELEN` is used to limit rule complexity, and `NTHREADS` is increased to optimize execution in the distributed CAS environment. The input table is slightly enlarged to better simulate a real-world use case with large data volumes.
Copied!
| 1 | /* Démarrer une nouvelle session CAS pour un contrôle explicite */ |
| 2 | options casport=5570 cashost='localhost'; |
| 3 | cas my_new_session sessopts=(caslib=casuser timeout=1800 locale="en_US"); |
| 4 | LIBNAME mycas cas; |
| 5 | |
| 6 | /* Création de la table de données d'avis (avec plus de données pour simuler une grande table) */ |
| 7 | DATA mycas.reviews_large; |
| 8 | INFILE DATALINES delimiter='|' missover; |
| 9 | LENGTH text $300 category $20; |
| 10 | INPUT text$ positive category$ did; |
| 11 | DATALINES; |
| 12 | This is the greatest phone ever! love it!|1|electronics|1 |
| 13 | The phone's battery life is too short and screen resolution is low.|0|electronics|2 |
| 14 | The screen resolution is low, but I love this tv.|1|electronics|3 |
| 15 | The movie itself is great and I like it, although the resolution is low.|1|movies|4 |
| 16 | The movie's story is boring and the acting is poor.|0|movies|5 |
| 17 | I watched this movie on tv, it's not good on a small screen. |0|movies|6 |
| 18 | watched the movie first and loved it, the book is even better!|1|books |7 |
| 19 | I like the story in this book, they should put it on screen.|1|books|8 |
| 20 | I love the author, but this book is a waste of time, don't buy it.|0|books|9 |
| 21 | Fantastic product, highly recommend!|1|electronics|10 |
| 22 | Battery drains too fast, very disappointed.|0|electronics|11 |
| 23 | Excellent display and sound quality.|1|electronics|12 |
| 24 | A captivating story, a must-watch!|1|movies|13 |
| 25 | Terrible plot and acting, waste of time.|0|movies|14 |
| 26 | Enjoyed reading every single page.|1|books|15 |
| 27 | Not worth the hype, prefer other titles.|0|books|16 |
| 28 | Best purchase this year!|1|electronics|17 |
| 29 | Worst experience ever, totally unreliable.|0|electronics|18 |
| 30 | Absolutely brilliant, a masterpiece.|1|movies|19 |
| 31 | Couldn't put it down, truly amazing.|1|books|20 |
| 32 | ; |
| 33 | run; |
| 34 | |
| 35 | /* Prétraitement du texte avec PROC TEXTMINE */ |
| 36 | proc textmine data=mycas.reviews_large; |
| 37 | doc_id |
| 38 | did; |
| 39 | var |
| 40 | text; |
| 41 | parse |
| 42 | nonoungroups |
| 43 | notagging |
| 44 | entities = none |
| 45 | outparent = mycas.reviews_bow_large |
| 46 | outterms = mycas.reviews_terms_large |
| 47 | reducef = 1; |
| 48 | run; |
| 49 | |
| 50 | /* Extraction des règles avec PROC BOOLRULE */ |
| 51 | proc boolrule |
| 52 | data = mycas.reviews_bow_large |
| 53 | docid = _document_ |
| 54 | termid = _termnum_ |
| 55 | docinfo = mycas.reviews_large |
| 56 | terminfo = mycas.reviews_terms_large |
| 57 | minsupports = 0.1 /* Support minimum de 10% */ |
| 58 | mpos = 0.5 /* Minimum de positivité de 50% */ |
| 59 | gpos = 0.5 /* Global positivité de 50% */ |
| 60 | maxrulelen = 2 /* Longueur maximale de la règle */ |
| 61 | nthreads = 8; /* Utiliser 8 threads pour le traitement */ |
| 62 | docinfo |
| 63 | id = did |
| 64 | targets = (positive); |
| 65 | terminfo |
| 66 | id = key |
| 67 | label = term; |
| 68 | output |
| 69 | ruleterms = mycas.ruleterms_cas |
| 70 | rules = mycas.rules_cas; |
| 71 | run; |
| 72 | |
| 73 | /* Affichage des règles générées */ |
| 74 | data rules_cas; |
| 75 | set mycas.rules_cas; |
| 76 | run; |
| 77 | |
| 78 | proc print data=rules_cas; |
| 79 | title 'Exemple Viya/CAS: Règles avec gestion de SESSION et performances'; |
| 80 | var target ruleid rule F1 precision recall; |
| 81 | run; |
| 82 | |
| 83 | /* Nettoyage et fin de session CAS */ |
| 84 | proc casutil; |
| 85 | droptable casdata='reviews_large' incaslib='mycas'; |
| 86 | droptable casdata='reviews_bow_large' incaslib='mycas'; |
| 87 | droptable casdata='reviews_terms_large' incaslib='mycas'; |
| 88 | droptable casdata='ruleterms_cas' incaslib='mycas'; |
| 89 | droptable casdata='rules_cas' incaslib='mycas'; |
| 90 | QUIT; |
| 91 | cas my_new_session terminate; |
| 92 |
This material is provided "as is" by We Are Cas. There are no warranties, expressed or implied, as to merchantability or fitness for a particular purpose regarding the materials or code contained herein. We Are Cas is not responsible for errors in this material as it now exists or will exist, nor does We Are Cas provide technical support for it.
Copyright Info : Copyright © SAS Institute Inc. All rights reserved.
