import streamlit as st import numpy as np import pandas as pd import seaborn as sns import matplotlib.pyplot as plt import plotly.express as px import nltk from nltk.tokenize import word_tokenize import wordcloud from wordcloud import WordCloud from sklearn.preprocessing import LabelEncoder from PIL import Image st.set_page_config( page_title='Sentiment Analysis of Amazon Customer Reviews', layout = 'wide', initial_sidebar_state='expanded' ) def run(): # title st.title('Sentiment Analysis of Amazon Customer Reviews') st.write('by Ahmad Luay Adnani') # sub header st.subheader ('Exploratory Data Analysis of the Dataset.') # Add Image image = Image.open('amazon.jpg') st.image(image,caption = 'illustration') # Description st.write('Amazon is one of the largest e-commerce platforms in the world, with millions of products available for purchase and billions of reviews submitted by customers. The reviews can vary in length, tone, and style, and often contain sarcasm, irony, or other forms of nuanced language. Sentiment analysis on Amazon reviews can **provide valuable insights for businesses**, including identifying common issues that customers face with their products or services, understanding the factors that drive customer satisfaction, and tracking changes in customer sentiment over time.') st.write('# Dataset') st.write('Dataset used is amazon review dataset from [kaggle]("https://www.kaggle.com/datasets/yacharki/amazon-reviews-for-sa-binary-negative-positive-csv").') # show dataframe df2 = pd.read_csv('dataset_20000_rows.csv') df2 = df2.drop(['Unnamed: 0'],axis=1) st.dataframe(df2) # add description of Dataset st.write('In this dataset, class 1 is the **negative review** and class 2 is the **positive review**') ### # create a copy of the dataframe df_eda = df2.copy() df_eda.class_index.replace({1:'Negative Review',2:'Positive Review'}, inplace=True) # Separating positive & negative review positive_review = df_eda[df_eda['class_index']=='Positive Review'] negative_review = df_eda[df_eda['class_index']=='Negative Review'] # Histogram and Boxplot based on user input st.write('# Exploratory Data Analysis') select_eda = st.selectbox('Select EDA : ', ('Type of Review','Example of Positive and Negative Review','Number of Words','WordCloud')) if select_eda == 'Type of Review': review = df_eda['class_index'].value_counts().to_frame().reset_index() fig = px.pie(review,values='class_index', names='index',color_discrete_sequence=['red','blue']) fig.update_layout(title_text = "Type of Review") st.plotly_chart(fig) st.write('Based on the table and visualization above, it can be seen that both negative and positive reviews consist of 10,000 reviews each.') elif select_eda == 'Example of Positive and Negative Review': # Print sample reviews pd.set_option('display.width', None) sample_negative_review = df_eda[df_eda['class_index']=='Negative Review'].sample(20) sample_positive_review = df_eda[df_eda['class_index']=='Positive Review'].sample(20) # Print Sample of Negative Review st.write('Example of Negative Reviews') st.write('-'*100) for i in range(0,20): st.write(sample_negative_review.iloc[i,2]) st.write('-'*100) # Print Sample of Positive Review st.write('Example of Positive Reviews') st.write('-'*100) for i in range(0,20): st.write(sample_positive_review.iloc[i,2]) st.write('-'*100) elif select_eda == 'Number of Words': # Count the number of words in each review df_eda['len_words'] = df_eda['review_text'].apply(lambda x: len(nltk.word_tokenize(x))) # Histogram plot for each review fig, ax =plt.subplots(1,2,figsize=(30,10)) sns.histplot(ax=ax[0],data=df_eda[df_eda['class_index'] == 'Positive Review']['len_words'],kde=True) ax[0].set_title('Positive Review') sns.histplot(ax=ax[1],data=df_eda[df_eda['class_index'] == 'Negative Review']['len_words'],kde=True) ax[1].set_title('Negative Review') st.pyplot(fig) st.write('-'*100) # Print Max and Average number of words st.write('The maximum number of words on each review is ', df_eda['len_words'].max()) st.write('The average number of words on each review is', df_eda['len_words'].mean()) st.write('-'*100) # Print Max and Average number of words on positive review st.write('The maximum number of words on positive review is ', df_eda[df_eda['class_index']=='Positive Review']['len_words'].max()) st.write('The average number of words on positive review is', df_eda[df_eda['class_index']=='Positive Review']['len_words'].mean()) st.write('-'*100) # Print Max and Average number of words on negative review st.write('The maximum number of words on negative review is ', df_eda[df_eda['class_index']=='Negative Review']['len_words'].max()) st.write('The average number of words on negative review is', df_eda[df_eda['class_index']=='Negative Review']['len_words'].mean()) st.write('-'*100) st.write('Based on the information above, it is known that **negative reviews have, on average, more words than positive reviews**. Based on my assumption, there could be various reasons why negative reviews have, on average, more words than positive reviews. Here are some possible explanations:') st.write("1. **Complex issues**: Negative reviews might involve more complex issues or problems with the product or service, which require more detailed explanations and examples.") st.write('2. **Emotional expression**: Negative reviews might include more emotional expression, such as frustration or disappointment, which can lead to more detailed and expressive language.') st.write('3. **Expectations**: Negative reviews might involve higher expectations from customers, leading them to provide more detailed feedback in order to articulate their disappointment or frustration.') st.write('4. *Personal experience**: Negative reviews might be based on a more personal experience, such as a defective product or poor customer service, which can lead to a more detailed and personalized account of the issue.') else: # Creating wordcloud text_positive = positive_review.review_text.values cloud_positive = WordCloud(max_words=50, background_color="white",width=2000,height=1000).generate(" ".join(text_positive)) # Showing wordcloud plt.figure(figsize=(15,10)) plt.axis('off') plt.title("Positive Review",fontsize=20) plt.imshow(cloud_positive) plt.show() st.set_option('deprecation.showPyplotGlobalUse', False) st.pyplot() # Creating wordcloud text_negative = negative_review.review_text.values cloud_negative = WordCloud(max_words=50, background_color="black",width=2000,height=1000).generate(" ".join(text_negative)) # Showing wordcloud plt.figure(figsize=(15,10)) plt.axis('off') plt.title("Negative Review",fontsize=20) plt.imshow(cloud_negative) plt.show() st.set_option('deprecation.showPyplotGlobalUse', False) st.pyplot() st.write('From the 2 visualizations above, we can obtain the following information:') st.write("1. **Book**, **movie**, **cd** and **album** are the products most frequently reviewed by amazon customers.") st.write('2. The most frequent positive words used by amazon customers included **good**, **great**, **love**, **best**, and **easy**. ') st.write('3. The most common negative words used by amazon customers included **bad**, **used** and **problem**.') if __name__ == '__main__': run()