RAGHU INSTITUTE OF TECHNOLOGY

(AUTONOMOUS)
(Approved by AICTE, Affiliated to JNTUK, Accredited by NBA & NAAC)
Department of Computer Science and Engineering
DAKAMARRI, VISAKHAPATNAM

TOURISM AND RESTAURANT RECOMMENDATION SYSTEM

BASED ON CUSTOMER REVIEWS THROUGH SENTIMENT
ANALYSIS.

Under the Guidance of: A Project Report by:

Mrs. T. Lakshmi Prasanna 213J1A05E0: PODURU CHAITANYA SAI
Associate Professor 213J1A05F1: ROKKAM ASHRITHA
Department Of CSE 213J1A05H6: V LAV
213J1A05I3: VASIPALLI NITHISH
 Abstract

• This project develops a novel recommendation system that leverages sentiment analysis of
customer reviews to suggest personalized tourism destinations and restaurants. The Dataset
for this project is derived from various websites through which customers book their
travelling or through which customers book their slot for Exquisite Restaurant Cuisine. The
system utilizes natural language processing (NLP) techniques and Machine Learning
Algorithms to extract sentiment scores, which are then used to rank and recommend
destinations and restaurants based on user preferences. The system provides personalized
recommendations, improves accuracy through sentiment analysis, enhances user experience,
and offers valuable insights for businesses. Along with visualising its results carefully, one of
its tasks will be to improve accuracy. This project contributes a new approach to tourism and
restaurant recommendation, a comprehensive evaluation framework, and a scalable system
for various applications.
Literature Survey
Sno Year Authors Org Title Outcome
.

1. 2024 Choirul Huda, Yaya IEEE Smart Tourism Achieved significant improvements in accuracy (MAE and
Heryadi, Lukas, and Recommender System RMSE), with CBCF models reducing errors by over 80%.
Widodo Budiharto Modeling Based on Increased the completeness of the user-item data from 38% to
Hybrid Technique and 100%, improving recommendation quality. Used advanced
Content Boosted techniques like CBCF, sentiment analysis, and demographic data
Collaborative Filtering to handle recommendations for new users effectively.

2 2021 Yong Qin, Xinxin Sprin Ranking Tourist Sentiments extracted accurately represent tourist preferences
Wang, Zeshui Xu ger Attractions through and reviews. Intuitionistic and hesitant fuzzy theories ensure
Online Reviews: A robust TA ranking. Adapts to incorporate both individual and
Novel Method with group preferences. Identifies weaknesses (e.g., cost or
Intuitionistic and environment) for TAs, enabling targeted improvements.
Hesitant Fuzzy
Information Based on
Sentiment Analysis
3 2022 Twil Ali, Bencharef Elsev Analyzing tourism Four key topics were identified: Jamaâ-el-Fna atmosphere,
Omar, Kaloun ier reviews using an LDA shopping experience, citizen behavior, and overall touristic
Soulaimane topic-based sentiment experience. Reviews were classified as positive, neutral, or
analysis approach negative, revealing a strong correlation between sentiment
scores from the algorithms and the TripAdvisor bubble rating
system. Most reviews indicated a positive sentiment overall. The
sentiment analysis models were benchmarked, with TextBlob
achieving 77.3% accuracy, VADER at 72.6%, and JST at 69.6%.
The study offered actionable insights into tourist feedback,
emphasizing areas needing improvement, such as pricing and
local behavior.
Sno Year Authors Org Title Outcome
.

4. 2023 Ram Krishn Mishra, Elsev Knowledge based topic Random Forest and Decision Tree classifiers were the
J Angel Arul Jothi, ier retrieval for most accurate for predicting star ratings, achieving
Siddhaling recommendations and accuracy levels up to 99.98%.The salience and valence
Urolagin, Kayan tourism promotions calculations provided actionable insights for restaurant
Irani managers, highlighting areas of strength and needed
improvements (e.g., improving service time or food
quality). The system successfully generated personalized
restaurant recommendations based on user preferences
and sentiment analysis, improving decision-making for
tourists.

5. 2021 XIAO ZHOU, IEEE Intelligent Tourism Experimental results demonstrate the system's ability to
MINGZHAN SU, Recommendation generate recommendations that align with tourists'
GUANGHUI FENG, Algorithm based on preferences. Tour routes generated by the system result in
AND XINGHAN Text Mining and MP higher tourist satisfaction compared to those from
ZHOU Nerve Cell Model of traditional algorithms. The algorithm outperforms traditional
Multivariate methods in terms of time and space complexity,
Transportation Modes demonstrating its efficiency. The system consistently
produces reliable outputs across various transportation
modes, showcasing its stability and adaptability.

6. 2022 N. Saraswathi, T. Sasi Elsev Improving the accuracy Achieved a sentiment classification accuracy of 94.7%,
Rooba, S. ier of sentiment analysis significantly outperforming baseline methods.
Chakaravarthi using a linguistic rule- Demonstrated improvements in precision, recall, and F-
based feature selection measure (specific numerical values not provided).
method in tourism Information Gain (IG) filter showed the best performance
reviews. among statistical methods. Random Forest (RF) excelled
in specific datasets, while Support Vector Machine (SVM)
performed strongly in others. Naive Bayes (NB) contributed
moderately to the ensemble. Combining statistical filters
and classifiers significantly improved prediction accuracy.
Proposed System

Fig.9: Proposed system

Implementation
• 1. Data Collection:

• Goal: Gather customer reviews for tourism and restaurants.

• Sources: Kaggle datasets and other external/manual sources.
• Process: Collect and consolidate relevant reviews.

• 2. Data Preprocessing
• Goal: Prepare the raw data for analysis.
• Steps:
• Remove Duplicates: Eliminate redundant reviews to ensure data quality.
• Handle Missing Values: Address incomplete or missing data points.

• 3. Exploratory Data Analysis (EDA)

• Goal: Understand the dataset through visualization and trend analysis.
• Steps:
• Visualize Sentiment Distributions: Plot sentiment scores or categories (e.g.,
positive, neutral, negative).
• Analyze Trends: Identify patterns in the data, such as popular restaurants,
common sentiments, or seasonal trends.
• 4. Modeling

• Goal: Implement sentiment analysis models for recommendation purposes.

• Approaches:
• Lexicon-Based Models: Use predefined sentiment dictionaries to analyze reviews.
• Transformer Models: Employ advanced machine learning models like BERT for more nuanced
sentiment classification.

• 5. Evaluation
• Goal: Assess the performance of the models.
• Steps:
• Analyze Misclassifications: Investigate incorrect predictions to improve model performance.

• 6. Insights & Conclusion

• Goal: Derive actionable insights and finalize recommendations.
• Steps:
• Summarize Findings: Report key outcomes of the analysis.
• Implications for Recommendations: Use insights to tailor recommendations for users.

• 7. Final Outcome:
• User Receives Tailored Recommendations: Deliver personalized restaurant or tourist
destination suggestions based on customer reviews and sentiment analysis.
Use case Diagram

Fig.10: Use case diagram

• This use case diagram outlines the interactions between three actors (Member, Guest, and
Administrator) and the system's functionalities.
• Actors
1. Member: Registered user with enhanced privileges.
2. Guest: User with limited functionality access.
3. Administrator: User responsible for system management.
• Use Cases
• Common to Member and Guest:
• Search Food, and Type of Tourism
• View Recommendations
• Exclusive to Administrator:
• Add/Delete Food and Restaurants
• Delete Members
• Edit Recommendations

• Key Observations:

• Shared Functionality: Core search and view features are shared between Members and Guests.
• Administrative Control: Administrators manage system data (e.g., food, restaurants, users).
• Differentiated Access: Members have additional privileges, such as saving recommendations.
• This diagram provides a clear overview of roles, access levels, and system functionalities.
Sequence diagram

Fig.11: Sequence diagram

• This sequence diagram illustrates the interaction flow between the User, Recommendation System, Database, and
Sentiment Analysis module to generate tailored recommendations.

Participants
1. User: Provides preferences and receives recommendations.
2. Recommendation System: Processes inputs, fetches data, and integrates sentiment scores.
3. Database: Stores data on restaurants, food, and tourism options.
4. Sentiment Analysis: Evaluates data to generate sentiment scores.

Sequence Flow

1. User Input: User specifies preferences (e.g., location, food, tourism type).
2. Data Retrieval: The system queries the database and retrieves relevant data iteratively for each user choice.
3. Sentiment Analysis: The retrieved data is analyzed to calculate sentiment scores (e.g., high or low).

4. Decision Making:
1. High Sentiment Scores: Best matches are recommended.
2. Low Sentiment Scores: Alternative suggestions are provided.

5. Final Output: The user receives personalized recommendations or alternatives.

• Key Observations
• Iterative Processing: Handles multiple inputs dynamically in a loop.
• Sentiment-Based Refinement: Sentiment analysis ensures recommendations are relevant and high-quality.
• Adaptive Decision Path: Recommendations adjust based on sentiment scores for a tailored user experience.
Activity diagram

Fig.12: Activity
diagram
• This activity diagram illustrates the step-by-step process of how the system processes user inputs
and generates recommendations.
• Key Components
1. Start: The process begins when the user interacts with the system.
2. User Input: The user specifies:
1. Food Choice: Preferred cuisine or dish.
2. Tourism Choice: Desired tourist activity or destination.
3. Recommendation System:
1. Processes inputs and queries the database for relevant options.
4. Database Interaction:
1. Retrieves data matching user preferences.
5. Sentiment Analysis:
• Calculates sentiment scores to evaluate the quality and relevance of options.
• Decision Points:
• Best Restaurant:
• Yes: Provides a restaurant recommendation.
• No: Indicates no suitable options.
• Best Tourist Destination:
• Yes: Recommends a tourist destination.
• No: Indicates no suitable options.
• End State
• The process concludes after recommendations (or fallback messages) are provided.

• Key Observations:
• Parallel Processing: Simultaneous evaluation of restaurants and destinations ensures efficiency.
• Fallback Handling: Robust paths for cases with no suitable recommendations.
• Sentiment Scores: Central to decision-making, improving recommendation quality.
Results

Ipywidget output for testing the system

• Selecting Mangalorean food for FOOD choice and
TEMPLE as place.

Testing Successful
• Gradio Web-App Interface for PC.

Recommendations for selected food and place

• Gradio Web-App Interface for Mobile Phones.

Recommendations for selected food and place

Conclusion
• The Recommendation System is a structured and user-centric platform that utilizes
sentiment analysis and a database-driven approach to provide personalized suggestions for
restaurants and tourist destinations. Here's the conclusion:

1. Personalization: The system takes into account user preferences for location, food, and
tourism types, ensuring tailored recommendations.
2. Efficiency: By leveraging a database and sentiment analysis, the system processes inputs
efficiently, providing meaningful and relevant suggestions.
3. Sentiment-Driven Decisions: The integration of sentiment analysis enhances the decision-
making process, ensuring that recommendations are based on quality and user satisfaction.
4. Scalability: This framework can be expanded to include additional parameters, such as
budget or user reviews, to further enhance the recommendation process.

• Overall, the system demonstrates a well-designed workflow that balances user input, data
processing, and quality output, making it a valuable tool for improving user experience in
tourism and dining contexts.