https://github.com/gui-sitton/detectnegativereviews

Create a model to classify reviews as positive and negative.
https://github.com/gui-sitton/detectnegativereviews

catboost logisticregression nltk nltk-python random-forest-classifier text-analysis text-classification tfidf-vectorizer xgboost

Last synced: about 2 months ago
JSON representation

Create a model to classify reviews as positive and negative.

• Host: GitHub
• URL: https://github.com/gui-sitton/detectnegativereviews
• Owner: Gui-Sitton
• Created: 2023-10-16T19:47:18.000Z (over 1 year ago)
• Default Branch: main
• Last Pushed: 2023-11-29T17:34:49.000Z (over 1 year ago)
• Last Synced: 2025-01-24T09:12:25.112Z (3 months ago)
• Topics: catboost, logisticregression, nltk, nltk-python, random-forest-classifier, text-analysis, text-classification, tfidf-vectorizer, xgboost
• Language: Jupyter Notebook
• Homepage:
• Size: 1.15 MB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# DetectNegativeReviews

## Intro

The Film Junky Union, a new community for classic film enthusiasts, is developing a system to filter and categorize film reviews. The aim is to train a model to automatically detect negative reviews. You will use a dataset of movie reviews from IMDB with polarity labeling to create a model to classify reviews as positive and negative. It will need to achieve an F1 value of at least 0.85.

### Data description

The data was provided by Andrew L. Maas, Raymond E. Daly, Peter T. Pham, Dan Huang, Andrew Y. Ng and Christopher Potts. (2011). Word learning vectors for sentiment analysis. The 49th Annual Meeting of the Association for Computational Linguistics (ACL 2011).

**Selected fields:**

review: the text of the review

pos: the goal, '0' for negative and '1' for positive

ds_part: 'train'/'test' for the training/testing part of the dataset, respectively

There are other fields in the dataset.

## Libraries used

import math

import re

import numpy as np

import pandas as pd

from sklearn.model_selection import train_test_split

import matplotlib

import matplotlib.pyplot as plt

import matplotlib.dates as mdates

import seaborn as sns

from nltk.corpus import stopwords as nltk_stopwords

from sklearn.feature_extraction.text import TfidfVectorizer

from sklearn.metrics import accuracy_score, f1_score

from tqdm.auto import tqdm

from sklearn.linear_model import LogisticRegression

import nltk

from nltk.corpus import stopwords as nltk_stopwords

nltk.download('stopwords')

from lightgbm import LGBMClassifier

from sklearn.ensemble import RandomForestClassifier

import xgboost as xgb

from catboost import CatBoostClassifier

## Conclusion

* Linear regression and LGBM were the best models, while the model 0 performed like a random model.

* Linear regression performed slightly better with the first type of text processing, normalization. With an accuracy of 88% and an f1 value of 0.88. The only difference was the APS value in the test set, from 0.95 with the first model to 0.94 with the second model using other text processing, which obtained a lower f1 value, also resulting in 0.87.

* In my reviews, all the models I tested performed as well as a random model, as I believe the dataframe was too small to work with and I didn't have enough examples, so it resulted in something like 50% accuracy for all the models.

# Detectar avaliações negativas

## Introdução

A Film Junky Union, uma nova comunidade para entusiastas de filmes clássicos, está desenvolvendo um sistema para filtrar e categorizar críticas de filmes. O objetivo é treinar um modelo para detectar automaticamente críticas negativas. Você usará um conjunto de dados de resenhas de filmes do IMDB com rotulagem de polaridade para criar um modelo para classificar as resenhas como positivas e negativas. Ele precisará atingir um valor F1 de pelo menos 0,85.

### Descrição dos dados

Os dados foram fornecidos por Andrew L. Maas, Raymond E. Daly, Peter T. Pham, Dan Huang, Andrew Y. Ng e Christopher Potts. (2011). Word learning vectors for sentiment analysis (Vetores de aprendizado de palavras para análise de sentimentos). 49ª Reunião Anual da Associação de Linguística Computacional (ACL 2011).

**Campos selecionados:**

revisão: o texto da revisão

pos: o objetivo, "0" para negativo e "1" para positivo

ds_part: 'train'/'test' para a parte de treinamento/teste do conjunto de dados, respectivamente

Há outros campos no conjunto de dados.

## Bibliotecas usadas

import math

import re

import numpy as np

import pandas as pd

from sklearn.model_selection import train_test_split

import matplotlib

import matplotlib.pyplot as plt

import matplotlib.dates as mdates

import seaborn as sns

from nltk.corpus import stopwords as nltk_stopwords

from sklearn.feature_extraction.text import TfidfVectorizer

from sklearn.metrics import accuracy_score, f1_score

from tqdm.auto import tqdm

from sklearn.linear_model import LogisticRegression

importar nltk

from nltk.corpus import stopwords as nltk_stopwords

nltk.download('stopwords')

from lightgbm import LGBMClassifier

from sklearn.ensemble import RandomForestClassifier

importar xgboost como xgb

from catboost import CatBoostClassifier

## Conclusão

* A regressão linear e o LGBM foram os melhores modelos, enquanto o modelo 0 funcionou como um modelo aleatório.

* A regressão linear teve um desempenho ligeiramente melhor com o primeiro tipo de processamento de texto, a normalização. Com uma precisão de 88% e um valor f1 de 0,88. A única diferença foi o valor APS no conjunto de teste, de 0,95 com o primeiro modelo para 0,94 com o segundo modelo usando outro processamento de texto, que obteve um valor f1 menor, também resultando em 0,87.

* Em minhas análises, todos os modelos que testei tiveram o mesmo desempenho de um modelo aleatório, pois acredito que o quadro de dados era muito pequeno para trabalhar e eu não tinha exemplos suficientes, o que resultou em algo como 50% de precisão para todos os modelos.