Smartphone Price Prediction in Big Data Environment

Table of Contents

1. Project Overview
2. Technologies Used
3. Architecture
4. Repository Structure
5. Software Requirements for Running the Project
6. How to Run
7. Dashboards
8. Acknowledgments
9. Conclusion
10. Contacts

1. Project Overview

This project aims to predict smartphone prices using a combination of batch and stream processing techniques in a Big Data environment. The architecture follows the Lambda Architecture pattern, providing both real-time and batch processing capabilities to users.

2. Technologies Used

• Ingestion Layer: Apache Kafka (message broker)
• Stream Layer: XGBoost (machine learning model), Apache HBase (real-time View)
• Batch Layer: Apache Spark (data processing framework), Apache Airflow (workflow orchestration), PostgreSQL (data warehouse (Batch View))
• Visualization: Spring Boot (web application framework), Power BI (interactive dashboards)

3. Architecture

• Here is the architecture :

The project architecture consists of five main layers: the ingestion layer, the batch layer, the stream layer, the serving layer and the visualization layer.

Ingestion Layer

• Apache Kafka: Utilized for real-time data ingestion from an API providing smartphone data.
  • Consumer: Collects data from the API and feeds it into the stream and batch layer.

Stream Layer

• Producer: A machine learning model developed using XGBoost to estimate smartphone prices. This model runs in real-time and stores predictions in a realtime view. (details about the model here )

Batch Layer

• HDFS: Data from the API is stored in HDFS as part of the data lake solution.
• PySpark: Performs data transformation on stored data using PySpark.
• Apache Airflow: Orchestrates the batch processing workflow.

Serving Layer

• Realtime View: Implemented using HBase to provide real-time access to predicted smartphone prices.
• Batch View: Transformed data is stored in PostgreSQL, as the data warehouse solution.

Visualization Layer

• Spring Boot Web Application: Provides a user interface to view real-time smartphone prices.
• Power BI Dashboard: Provides batch users with a visualization of processed data.

4. Repository Structure

The repository is organized as follows:

Big-Data-Project:.
|   README.md
|
+---images
|       architecture.png
|       dashboard_phone.png
|       run_web_app.png
|       spring_boot_web_app.png
|
\---Main
    |   commands.sh
    |   Dashboard.pbix
    |
    +---.idea
    |       workspace.xml
    |
    +---Lambda
    |   |   docker-compose.yaml
    |   |   producer.py
    |   |   transform.py
    |   |
    |   +---.idea
    |   |   |   .gitignore
    |   |   |   .name
    |   |   |   misc.xml
    |   |   |   modules.xml
    |   |   |   price prediction (big data envirnment).iml
    |   |   |   vcs.xml
    |   |   |   workspace.xml
    |   |   |
    |   |   \---inspectionProfiles
    |   |           profiles_settings.xml
    |   |
    |   +---Batch_layer
    |   |   |   batch_layer.py
    |   |   |   batch_pipeline.py
    |   |   |   HDFS_consumer.py
    |   |   |   put_data_hdfs.py
    |   |   |   save_data_postgresql.py
    |   |   |   spark_tranformation.py
    |   |   |   __init__.py
    |   |   |
    |   |   +---dags
    |   |   |       syc_with_Airflow.py
    |   |   |       __init__.py
    |   |   |
    |   |   \---__pycache__
    |   |           batch_layer.cpython-310.pyc
    |   |           HDFS_consumer.cpython-310.pyc
    |   |           put_data_hdfs.cpython-310.pyc
    |   |           save_data_postgresql.cpython-310.pyc
    |   |           spark_tranformation.cpython-310.pyc
    |   |           __init__.cpython-310.pyc
    |   |
    |   +---ML_operations
    |   |   |   xgb_model.pkl
    |   |   |
    |   |   \---__pycache__
    |   +---real_time_web_app(Flask)
    |   |   |   app.py
    |   |   |   get_Data_from_hbase.py
    |   |   |
    |   |   +---static
    |   |   |   +---css
    |   |   |   |       style.css
    |   |   |   |
    |   |   |   \---js
    |   |   |           script.js
    |   |   |
    |   |   +---templates
    |   |   |       index.html
    |   |   |
    |   |   \---__pycache__
    |   |           get_Data_from_hbase.cpython-310.pyc
    |   |
    |   +---Stream_data
    |   |   |   stream_data.csv
    |   |   |   stream_data.py
    |   |   |
    |   |   \---__pycache__
    |   +---Stream_layer
    |   |       insert_data_hbase.py
    |   |       ML_consumer.py
    |   |       stream_pipeline.py
    |   |       __init__.py
    |   |
    |   \---__pycache__
    |           producer.cpython-310.pyc
    |           transform.cpython-310.pyc
    |
    \---real_time_app(Spring boot)
        |   .classpath
        |   .gitignore
        |   .project
        |   HELP.md
        |   mvnw
        |   mvnw.cmd
        |   pom.xml
        |
        +---.mvn
        |   \---wrapper
        |           maven-wrapper.jar
        |           maven-wrapper.properties
        |
        +---.settings
        |       org.eclipse.core.resources.prefs
        |       org.eclipse.jdt.core.prefs
        |       org.eclipse.m2e.core.prefs
        |
        +---src
        |   +---main
        |   |   +---java
        |   |   |   \---com
        |   |   |       \---example
        |   |   |           \---demo
        |   |   |               |   RealTimeAppApplication.java
        |   |   |               |
        |   |   |               +---controller
        |   |   |               |       IndexController.java
        |   |   |               |
        |   |   |               \---service
        |   |   |                       HbaseService.java
        |   |   |
        |   |   \---resources
        |   |       |   application.properties
        |   |       |
        |   |       +---static
        |   |       |   +---css
        |   |       |   |       style.css
        |   |       |   |
        |   |       |   \---js
        |   |       |           script.js
        |   |       |
        |   |       \---templates
        |   |               index.html
        |   |
        |   \---test
        |       \---java
        |           \---com
        |               \---example
        |                   \---demo
        |                           RealTimeAppApplicationTests.java
        |
        \---target
            +---classes
            |   |   application.properties
            |   |
            |   +---com
            |   |   \---example
            |   |       \---demo
            |   |           |   RealTimeAppApplication.class
            |   |           |
            |   |           +---controller
            |   |           |       IndexController.class
            |   |           |
            |   |           \---service
            |   |                   HbaseService.class
            |   |
            |   +---META-INF
            |   |   |   MANIFEST.MF
            |   |   |
            |   |   \---maven
            |   |       \---com.example
            |   |           \---real_time_app
            |   |                   pom.properties
            |   |                   pom.xml
            |   |
            |   +---static
            |   |   +---css
            |   |   |       style.css
            |   |   |
            |   |   \---js
            |   |           script.js
            |   |
            |   \---templates
            |           index.html
            |
            \---test-classes
                \---com
                    \---example
                        \---demo
                                RealTimeAppApplicationTests.class

5. Software Requirements for Running the Project

This project requires the following software to be installed and configured on your system:

Big Data Stack:

• Apache Kafka (version 2.6.0)
• Apache HBase (version 1.2.6)
• Apache Hadoop (version 2.7.0)
• Apache Spark (version 3.3.4)
• PostgreSQL database

Programming Languages and Frameworks:

• Python (version 3.10.x or later)
• Java 17 (or compatible version)
• Spring Boot

Machine Learning Library:

• XGBoost

Additional Tools:

• Apache Airflow
• Power BI Desktop

By installing and configuring these tools, you will have the necessary environment to run this project and leverage its real-time and batch processing capabilities for smartphone price prediction and analysis.

6. How to Run

To set up and run the project locally, follow these steps:

• Clone the repository:

git clone https://github.com/aymane-maghouti/Big-Data-Project

1. Stream Layer

• Start Apache zookeeper

zookeeper-server-start.bat C:/kafka_2.13_2.6.0/config/zookeeper.properties

• Start Kafka server

kafka-server-start.bat C:/kafka_2.13_2.6.0/config/server.properties

• Create Kafka topic

kafka-topics.bat --create --topic smartphoneTopic --bootstrap-server localhost:9092

• Run the kafka producer

kafka-console-producer.bat --topic smartphoneTopic --bootstrap-server localhost:9092

• Run the kafka consumer

kafka-console-consumer.bat --topic smartphoneTopic --from-beginning --bootstrap-server localhost:9092

• Start HDFS and yarn (start-all or start-dfs and start-yarn)

start-all  

• Start Hbase

start-hbase  

• Run thrift server (for Hbase)

hbase thrift start

after all this run stream_pipeline.py script.

and then open the spring boot appliation in your idea and run it (you can access to the web app locally on localhost:8081/)

---

note that there is another version of the web app developed using Flask micro-framework(watch the demo video for mor details)

2. Batch Layer

• Start the Apache Airflow instance:

docker-compose up -d

Access the Apache Airflow web UI (localhost:8080) and run the DAG

• Start Apache Spark

spark-shell

• Start Apache zookeeper

zookeeper-server-start.bat C:/kafka_2.13_2.6.0/config/zookeeper.properties

• Start Kafka server

kafka-server-start.bat C:/kafka_2.13_2.6.0/config/server.properties

• Run the kafka producer

kafka-console-producer.bat --topic smartphoneTopic --bootstrap-server localhost:9092

• Run the kafka consumer

kafka-console-consumer.bat --topic smartphoneTopic --from-beginning --bootstrap-server localhost:9092

• Run HDFS and yarn (start-all or start-dfs and start-yarn)

start-all  

• Open power BI file dashboard.pbix attached with this project

after all this run syc_with_Airflow.py script.

7. Dashboards

This project utilizes two dashboards to visualize smartphone price predictions and historical data:

1. Real-Time Dashboard (Spring Boot Application):

• This dashboard is built using a Spring Boot web application.
• It displays the predicted price of smartphones in real-time.
• Users can access this dashboard through a web interface.

Here is the UI of th Spring Boot web application:

2. Batch Dashboard (Power BI):

• This dashboard leverages Power BI for interactive data exploration.
• It provides insights into historical smartphone price trends.
• This dashboard is designed for batch users interested in historical analysis.

Here is the Dashboard created in Power BI:

8. Acknowledgments

• Special thanks to the open-source communities behind Python, Kafka, HDFS , Spark,Hbase,Spring Bootand Airflow

9. Conclusion

• This big data architecture effectively predicts smartphone prices in real-time and provides historical analysis capabilities. The Lambda architecture facilitates efficient stream processing for real-time predictions using XGBoost and HBase, while Apache Airflow orchestrates batch processing with Spark to populate the PostgreSQL data warehouse for historical insights. This solution empowers real-time and batch users with valuable price information, enabling data-driven decision-making.

you can watch the demo video here

10. Contacts

For any inquiries or further information, please contact:

• Name: Aymane Maghouti
• Email: [email protected]
• LinkedIn: Aymane Maghouti