Enhancing Travel Package Decision-Making Using Time Series-Based Price

Forecasting and Alert Generation: A Case Study on SME Travel Portal

1.K.Midhun Kumar, Dr.Zafar Ali Khan N
Department of Computer Science & Engineering, School of Engineering & Technology,
Presidency University Bangalore, India

Abstract current price, without any guidance on whether it

This research presents a price prediction might rise or fall in the near future.
and alert generation module developed as an
enhancement to an existing travel website project. This paper presents a machine learning-based
The module uses time series forecasting, enhancement developed for an academic travel
specifically the Facebook Prophet model, to predict platform project (SME Travels), which forecasts
future prices of travel packages based on historical package prices and alerts users accordingly. The
pricing trends. Simulated data was used to train the objective is to support smarter travel planning by
model, and results were integrated into the website leveraging data-driven insights.
interface to provide users with simple booking
The SME Travel project was developed as a
advice like “book now” or “wait for a price drop.”
web-based platform to manage and display tour
The system outputs forecast charts and generates
packages for users. The site allows package
JSON-based alert messages for display on package
creation, itinerary updates, and package
pages. This approach introduces predictive
categorization but does not track or analyze price
intelligence to the platform, helping users plan more
behavior over time. Travel prices often fluctuate
cost-effectively while showcasing the application of
due to seasonal demand, special events, and
machine learning in real-world tourism systems.
administrative changes, yet users lack visibility into
these patterns.
Keywords
To bridge this gap, a time series forecasting
Travel Package, Time Series Forecasting, Facebook
system was independently developed and added as a
Prophet, Price Prediction, User Alert, Machine
research enhancement. It predicts price changes
Learning, Tourism, Smart Booking
using a pre-trained model and outputs booking
suggestions that are embedded into the frontend of
I. INTRODUCTION the platform.

In recent years, travel websites have evolved

from basic listing platforms to full-fledged online II. LITERATURE REVIEW
tour management systems. However, while many
2.1 Traditional Travel Systems
platforms offer package information, schedules, and
bookings, few provide intelligent insights into how Many existing travel websites focus primarily on
package prices might change over time. Users are listing packages and allowing user bookings, with
often forced to make decisions based only on the prices manually entered and updated by
administrators. These platforms typically lack
intelligent features for forecasting or guiding user The proposed system for price forecasting and alert
booking decisions based on price trends [1]. generation for travel packages leverages data
analysis and machine learning techniques to provide
2.2 AI and ML in Tourism :
real-time insights into price fluctuations. The
Artificial Intelligence has found increasing use in methodology is structured in the following key
tourism applications such as itinerary planning, steps:
chatbots, destination recommendations, and
behavior-based personalization. Several systems
leverage machine learning to suggest places,
optimize routes, and provide virtual guidance [2][3].
However, dynamic pricing prediction —
particularly using time series — is still largely
unexplored in user-facing travel systems.

2.3 Time Series Forecasting in Technology :

Time series forecasting has been widely used in
fields like stock market prediction, energy
consumption estimation, weather forecasting, and
retail sales analysis. Techniques such as ARIMA,
LSTM, and Prophet have been proven effective.
Prophet, in particular, is known for its ease of use,
support for automatic seasonality detection, and
strong performance on business data with missing
values or irregular updates [4][5].
figure - 1

2.4 Research Gaps: 3.1 Data Simulation and Preparation

While AI-based personalization is becoming
Objective: Create a simulated historical dataset to
common in travel systems, predictive analytics
mimic real-world price fluctuations.
focused on future package pricing is still
underutilized. There is a lack of systems that
3.1.1 Data Collection:
proactively alert users when prices are expected to
Collect initial data for various travel packages,
rise or fall. This research addresses this gap by
including attributes such as Package_ID (unique
proposing a lightweight, modular system using
identifier for each package), Price (travel package
machine learning for intelligent price forecasting
price), and Date of Update (date when the price was
and alert generation integrated into a travel booking
last updated).
platform.
III. PROPOSED METHODOLOGY 3.1.2 Price Fluctuation Simulation:
Introduce artificial fluctuations to the price data for
the past few months to reflect typical market
changes. These fluctuations can be either random or 3.3 Forecasting with Prophet
based on historical trends (seasonality, special
events, etc.). Objective: Use Prophet to generate forecasts for
future price trends.
3.1.3 Enhancement:
In addition to manual fluctuations, consider adding 3.3.1 Model Initialization:
external factors such as holidays or promotional For each travel package, initialize a Prophet model
discounts, which could impact pricing. with the desired parameters (e.g., seasonalities,
holidays, etc.) to capture underlying trends and
seasonality.

3.3.2 Model Training:

3.2 Data Preprocessing and Transformation Fit the Prophet model using the historical price data
(ds and y columns).
Objective: Transform raw data into a format
suitable for forecasting. 3.3.3 Forecasting:
Generate forecasts for the next 90 days (future
3.2.1 Data Import: price predictions).
Load the dataset using Python and the Pandas
library. 3.3.4 Extract Predictions:
Extract the predicted values (yhat) for the future
3.2.2 Data Cleaning:
dates.
Handle missing values, outliers, and ensure
consistency in the format (e.g., checking for any
3.3.5 Enhancement:
erroneous dates or prices).
Fine-tune the model to account for special
conditions like major events or promotions that
3.2.3 Feature Engineering:
could cause significant price changes. This can
Create necessary features like ds, the date column involve adding custom holiday effects or adjusting
representing the time dimension, and y, the price seasonalities.
column representing the target variable (travel
package price). 3.4 Trend Evaluation and Alert Generation

3.2.4 Grouping Data: Objective: Evaluate the forecast trends and

Group the dataset by Package_ID to forecast each generate dynamic price alerts.
package's price separately, ensuring personalized
forecasting models for each travel package. 3.4.1 Trend Comparison:
Compare the first predicted price (yhat[0]) and the
3.2.5 Enhancement: last predicted price (yhat[-1]) for each package.
Implement a time-based cross-validation technique
to check the robustness of the dataset and train-test 3.4.2 Alert Generation:
splits, ensuring better generalization of the models. If the forecast shows a price increase, generate an
alert message such as:
 “Prices may rise soon. Book now.” JSON file containing forecast data and alert
If the forecast shows a price decrease, generate an messages. Dynamically display the appropriate
alert message such as: message on the travel package detail page based on
“Price drop expected. Consider waiting.” the forecast trend (e.g., price increase or decrease).

3.4.3 Enhancement: 3.6.2 Visualization:

Implement a threshold for change, such that the Use Chart.js or a similar JavaScript library to
alert is only triggered if the price change exceeds a render an interactive line chart on the frontend,
certain percentage (e.g., 5% increase or decrease). displaying the predicted prices for the next 90 days.
This helps avoid generating too many trivial alerts. Alternatively, display the static PNG chart if real-
time rendering is not necessary.

3.5 Output Generation and Storage 3.6.3 Enhancement:

Consider adding interactivity to the chart (e.g.,
Objective: Save the forecast data and alerts for later
hover-over tooltips showing exact values for
use in the frontend system.
specific dates, zoom-in/zoom-out functionality for
different time spans, or comparison with actual
3.5.1 Data Output:
prices).
Store the forecast results in a JSON file containing
predicted price values (yhat) and the generated
alert messages.
IV. IMPLEMENTATION & RESULT
3.5.2 Visual Output:
Store the forecast trend visually in a PNG chart, 4.1 Overview of SME Travel Website
showing a time-series line chart of the predicted
The SME Travel website is a web-based tour
prices over the 90-day period.
package management system designed to offer a
range of domestic and international travel
3.5.3 Enhancement:
experiences. The application was developed using
Include additional metadata in the JSON output,
ASP.NET Web Forms for the backend and
such as confidence intervals (upper and lower
HTML, CSS, and JavaScript for the frontend. The
bounds of predictions), to provide a range of price
system was hosted locally using Internet
expectations for better decision-making.
Information Services (IIS) during development.
3.6 Frontend Integration and User The platform consists of two major components:
Interaction
● Admin Panel:
Objective: Integrate the forecasting model output The admin interface enables package
with the frontend system to display real-time alerts creation and editing. Admins can update
and visual trends. destination details, pricing, itineraries, and
assign service icons (such as hotel, food,
3.6.1 Dynamic Alert Display: transport, etc.). Packages can also be
Use JavaScript (e.g., fetch API) to fetch the categorized into various tour types including
 honeymoon, group, adventure, and ○ An alert message giving a booking
pilgrimage tours. suggestion (e.g., “Price may rise
soon. Book now.”)
● User Interface:
The user-facing interface displays available ○ A trend chart visualized using
travel packages in a categorized layout. Chart.js or a pre-generated image
Each package links to a detail page
containing the itinerary, highlights, and ● User Interaction:
pricing. The system does not include user The alert appears beneath the package price,
account features; users can directly browse and the line chart is embedded in the detail
the site without registration. section. Users can interpret upcoming price
movements before making booking
decisions.
This setup provides a clean and organized tour
browsing experience while allowing admins full
control over content management. This integration adds a lightweight but intelligent
layer to the system, providing data-driven insights
4.2 Integration of the Predictive System without altering the existing platform structure.

The predictive price forecasting feature was

implemented as a modular extension to the SME
Travel platform. It operates independently of the
backend and admin panel, allowing seamless
integration without modifying the database or
existing application logic.

The implementation flow is as follows:

● Model Output Placement:

After the model forecasts price trends for
the next 90 days, the outputs — including a
JSON file (with prediction values and alert
messages) and a line chart image — are Figure - 2
stored in a /static/predictions/ folder inside
the project. 4.3 System Architecture

● Frontend Fetch Logic: The system architecture for the proposed predictive
On the package detail page, a JavaScript feature is modular and layered over the existing
function retrieves the corresponding JSON SME Travel platform without disrupting its backend
file based on the package ID. It then extracts logic. The architecture consists of five key stages:
and displays:
1. Admin Data Update (Manual)
The admin updates travel package prices
manually through the existing web portal.
 These updates simulate real-world price ● Frontend: HTML, CSS, JavaScript, and
changes over time. Chart.js for rendering the chart in the
browser.
2. Data Preparation and Model Training
Historical pricing data is structured and ● Backend (Static Integration): No
saved as CSV files. A Python script using modifications were made to the existing
the Facebook Prophet model trains on this ASP.NET backend. The predictive module
data for each Package_ID. runs separately.

3. Prediction and Output Generation ● Machine Learning Layer: Implemented in

The model forecasts prices for a 90-day Python using Pandas for data handling and
period. For each package, it generates: Facebook Prophet for time series
forecasting.
○ A JSON file containing future price
values and an alert message ● Output Handling: JSON files are used to
pass data and alerts to the frontend, while
○ An optional line chart image PNG charts provide visual representations of
showing the price trend trends.

4. Static Storage ● Hosting & Access: All prediction outputs

The generated files are stored in a local are placed in a static directory within the
/static/predictions/ folder accessible by the project’s public structure to be accessed by
frontend. JavaScript without requiring server-side
scripts.
5. Frontend Integration
JavaScript on the package detail page
dynamically fetches the corresponding V. OUTCOMES
JSON file and:
This research focuses on extending the functionality
○ Displays the alert message of a traditional travel booking system by
incorporating a predictive model that forecasts
○ Renders the price trend chart using package prices and offers booking suggestions. The
Chart.js system is built as an independent module and
integrated into the frontend interface without
disrupting existing backend logic. The following
This layered setup ensures the intelligent system
key objectives were defined to guide the
works independently, requiring no real-time API or
development:
database modification.

4.4 Technology Stack

Simulating Realistic Pricing Behavior
The following technologies were used across the To create a structured dataset that mimics how
system: travel package prices change over time due to
factors like seasonal demand, holidays, or internal
business strategies. This dataset acts as the real-world price fluctuations and leveraging
foundation for time series forecasting. historical data, the system can generate reliable
forecasts for the next 90 days, offering valuable
Applying Time Series Forecasting Techniques insights for both customers and administrators.
To implement a forecasting model using the
Facebook Prophet library in Python, enabling the The dynamic alert system ensures that users are
prediction of future travel package prices over a 90- notified in real-time about price trends, empowering
day period based on historical price data. them to make informed decisions. With seamless
integration into the frontend, the system provides
Trend Detection and Message Generation not only forecast data but also visual
To evaluate the trend of predicted prices and representations, such as charts and alerts, enhancing
generate relevant alert messages. For example, if the user experience.
prices are expected to rise, users receive a prompt to
book early; if a drop is predicted, users are Furthermore, the modular structure of the
encouraged to wait. methodology allows for future enhancements, such
as incorporating external factors influencing prices
Frontend Data Integration or optimizing the model with custom seasonal
To display the model’s predictions directly on the adjustments. This system, when fully implemented,
travel package detail page. This involves fetching will provide a significant edge in the competitive
JSON-based output and visualizing forecasted data travel industry by enabling smarter, data-driven
using JavaScript and Chart.js without altering the pricing strategies and improving customer
original backend. satisfaction.

Improving User Decision-Making

To support users in making smarter, cost-effective ACKNOWLEDGMENT
travel decisions by giving them insights into future I would like to acknowledge the support of
pricing trends, reducing uncertainty in booking Presidency University for providing resources and
timing. facilitating this research project. We are also
grateful to the university librarians, professors, and
Demonstrating the Role of AI in Web Platforms
research assistants for their assistance.
To showcase how AI and machine learning
techniques can be integrated into real-world web REFERENCES
applications for enhancing usability, interactivity, 1. [1] Vaidya, R., & Sahu, R. (2020). Tourism
and decision support in the tourism sector. Management System Using Web
Technologies. International Journal of
Computer Applications, 175(23), 12–15.
VII. CONCLUSION
2. [2] Tussyadiah, I., & Miller, G. (2019). Role
The proposed methodology for price forecasting of AI and chatbots in travel: User
and alert generation for travel packages combines interaction in automated services. Journal of
advanced data preprocessing techniques with Travel Research.
machine learning tools like Prophet to provide
accurate and timely price predictions. By simulating
3. [3] Fadda, S., & Sorace, S. (2021). AI Tourism Management System,” Volume 12,
Applications in Tourism: A Review. Tourism Issue 4, April 2024.
Review, 76(4), 857–869.

4. [4] Taylor, S. J., & Letham, B. (2018).

Forecasting at scale. The American
Statistician, 72(1), 37–45.

5. [5] Hyndman, R. J., & Athanasopoulos, G.

(2018). Forecasting: Principles and
Practice (2nd ed.). OTexts.
6. https://www.kaggle.com/code/prashant111/
tutorial-time-series-forecasting-with-prophet
- Kaggle Prophet
7. Merari, C.1 and Ibrahim, J. I., “A Website:
A Way to Promote the Products and
Tourism,” SK No. 230/E/KPT/2022, English
Department, Faculty of Letters, Petra
Christian University.
8. Asit Joshi, Ayush Choudhary, Deepakshi
Choudhary, Deependra Singh Parihar,
“Travel and Tourism Management System,”
Volume: 04/Issue: 11/November-2022.
9. De Silva, W.A.C.Pabasara, Rimas M.J.M,
Kumari J.M.D, Kumari A.A.M, Inkithai M,
Kapilan K, “Tourism Management Web
Implementation,” May 2023.
10. Gonde Neha Siddhareddi, Satav Divya
Tanaji, Nagargoje Krushnakant Rajendra,
Khedkar Vishwjeet Adinath, “Small Tour
Travel Planner,” Volume 8, Issue 4, April
2020 | ISSN: 2320-2882, November 2023.
11. Prof. Harishchandra Maurya, Kiran Kailash
Guthale, Roshani Brahmadeo Kolekar, Smit
Milind Shirke, Amey Hanuman Jejurkar,
“Trip4u: Responsive Tourism Website for
Tour and Travel Management,” Volume 3,
Issue 2, November 2023.
12. Jayesh Dhole, Harsh Pandey, Ayaan Sheikh,
Vaishnavi Sune, Prof. A. V. Sable, “Design
and Implementation of a Web-Based