10.sentiment analysis of customer product reviews using machine learni
Download as DOCX, PDF0 likes304 views
The document details a project on sentiment analysis of Amazon product reviews using machine learning algorithms including SVM, decision tree, and Naive Bayes, where SVM achieved the best accuracy. It describes the steps involved in the project such as data collection, preprocessing, feature extraction with TF-IDF, and the training of the algorithms. Finally, it provides insight into how sentiment predictions are made and the visual representation of algorithm accuracy.
10.sentiment analysis of customer product reviews using machine learni
- 1. Venkat Java Projects Mobile:+91 9966499110 Visit:www.venkatjavaprojects.com Email:venkatjavaprojects@gmail.com SentimentAnalysisof CustomerProductReviewsUsingMachine Learning In this project author is detecting sentiments from amazon reviews by using various machine learning algorithms such as SVM, Decision Tree and Naïve Bayes. In all 3 algorithms SVMis giving better accuracy and to train this algorithms author has used AMAZON reviews dataset and this datasetissavedinside ‘Amazon_Reviews_dataset’folder.Below screenshotshow example reviews from dataset In above datasetfirstrowcontainscolumnnamesandremainingrowscontainsdatasetvaluesandin above datasetfirstcolumncontainssentimentvaluesfrom1to 5 and itsassociatedwitheachreview and we will use above dataset to train all 3 machine learning algorithms. To implement this project author has used following modules 1) Data Collection:Using this module we will upload AMAZON reviews dataset to application 2) Data Preprocessing:usingthismodule we will readall reviewsandthenremove stop words, special symbols, punctuation and numeric data from all reviews and after applying Preprocessing we will extract features from all reviews. 3) Features Extraction: here we will apply TF-IDF (term frequency Inverse Document Frequency) algorithmtoconvertstringreviewsintonumericvector.Eachword count will be put in vector in place of words. 4) Run SVM Algorithm: We will apply SVMalgorithm on TF-IDF vector to train SVMalgorithm and then we apply test data on SVMtrained model to calculate SVMprediction accuracy 5) Run Naïve Bayes Algorithm: We will apply Naïve Bayes algorithm on TF-IDF vector to train Naïve Bayes algorithm and then we apply test data on Naïve Bayes trained model to calculate Naïve Bayes prediction accuracy
- 2. Venkat Java Projects Mobile:+91 9966499110 Visit:www.venkatjavaprojects.com Email:venkatjavaprojects@gmail.com 6) Run DecisionTree Algorithm:We will applyDecisionTree algorithmonTF-IDFvector to train Decision Tree algorithm and then we apply test data on Decision Tree trained model to calculate Decision Tree prediction accuracy 7) DetectSentimentfromTestReviews:Usingthismodulewe willuploadtestreviewsandthen ML algorithm will predict sentiment for each review and in below test reviews dataset we can see there is no sentiment value and ML will predict sentiment for each test value In above test data we have only test reviews and by applying ML trained model on above test data we can predict sentiment label. SCREEN SHOTS To run project double click on ‘run.bat’ file to get below screen
- 3. Venkat Java Projects Mobile:+91 9966499110 Visit:www.venkatjavaprojects.com Email:venkatjavaprojects@gmail.com In above screen click on ‘Upload Amazon Reviews Dataset’ button to upload dataset In above screenwe are selectinganduploading‘Amazon.csv’file and then click on ‘Open’ button to load dataset and to get below screen
- 4. Venkat Java Projects Mobile:+91 9966499110 Visit:www.venkatjavaprojects.com Email:venkatjavaprojects@gmail.com In above screen dataset loaded and now click on ‘Preprocess Dataset’ button to read all reviews from dataset and then apply Preprocess steps to get below screen In above black console we can see application read all reviews from dataset and then generate below TF-IDF vector
- 5. Venkat Java Projects Mobile:+91 9966499110 Visit:www.venkatjavaprojects.com Email:venkatjavaprojects@gmail.com In above screen in text area we can see application extract all words from reviews and then put in top line of above testareaand inremainingrowsif thatword appearthenitput average countvalue of that word and if word not appear then 0 will put. In above screen vector generated and I am showing few records from that vector. In that vector total reviews are 573 and all reviews contains total 2361 unique words.Nowvectorisready and now click on ‘Run SVMAlgorithm’ button to train SVMwith above vector
- 6. Venkat Java Projects Mobile:+91 9966499110 Visit:www.venkatjavaprojects.com Email:venkatjavaprojects@gmail.com In above screen with SVM we got 82% accuracy and now click on Naïve Bayes and Decision tree button to get their accuracy In above screenwithall 3 algorithms SVMgave better prediction accuracy and now click on ‘Detect Sentiment from Test Reviews’ button to upload test reviews In above screenselectinganduploading ‘test.csv’ file and then click on ‘Open’ button to get below prediction result
- 7. Venkat Java Projects Mobile:+91 9966499110 Visit:www.venkatjavaprojects.com Email:venkatjavaprojects@gmail.com In above screenfirstIam displayingreviewsfromuploadedtestfileandthenpredictingpositive and negative sentiment for each review and you can scroll down above text area to get all outputs In above screen we can see sentiment prediction result for all reviews and now click on ‘Accuracy Graph’ button to get below graph
- 8. Venkat Java Projects Mobile:+91 9966499110 Visit:www.venkatjavaprojects.com Email:venkatjavaprojects@gmail.com In above graphx-axisrepresentsalgorithmname andy-axisrepresents accuracy of those algorithms and in all 3 algorithms SVMgot higher accuracy