Data piling up is meaningless if nobody reads it. We help turn it into numbers you can use to make decisions.
This page is for Sentiment Analysis of Indonesian Product Reviews with TF-IDF, Naive Bayes, and SVM built with RapidMiner. This project classifies 2,400 fictional Indonesian product reviews as positive or negative. Text is turned into word attributes with TF-IDF weighting, Naive Bayes and SVM are compared, and the most influential words are shown for the marketing team.
What You Receive
- A ready-to-run RapidMiner process (.rmp), with operator settings you can open and modify
- A cleaned dataset: duplicates, messy formats, and missing values are tidied first
- Analysis results as CSV/Excel tables, for example svm accuracy (test data)
- A managerial interpretation report in language non-technical readers understand, with recommended actions
- A session explaining the results and a short guide to re-running the process
Operators Used
Read CSV, Remap Binominals, Generate Attributes, Select Attributes, Generate TFIDF, Replace Missing Values, Split Data, Multiply, Weight by Information Gain, Naive Bayes, Support Vector Machine (LibSVM), Apply Model, Performance (Binominal Classification), Performance to Data, Weights to Data, Write CSV.
Example Results from the Portfolio Project
- SVM accuracy (test data): 92,1%
- Naive Bayes accuracy (test data): 91,8%
- SVM F-measure: 0,927
- SVM / Naive Bayes AUC: 0,923 / 0,895
- Most influential word: menyesal
How the Process Works
- Question mapping — we first make sure which decision you want to take from this data
- Data audit — we check the completeness and quality of the data before analysis
- Processing and modeling — the process runs in RapidMiner, and assumptions and parameters are documented openly
- Joint review — results are discussed and the direction of the analysis can be adjusted
- Handover — the process file, the cleaned dataset, and the interpretation report



