ABSTRACT
This project presents the development of an intelligent sweetness prediction system using Machine Learning and a Flask-based web application for real-time monitoring and analysis. The main aim of the system is to predict the sweetness level of a sample based on live sensor data collected from an external API source. In many food processing and agricultural industries, sweetness is considered an important quality parameter, and manual measurement methods can be time-consuming, inconsistent, and less efficient. To overcome this problem, the proposed system automates the prediction process by using a trained machine learning model that can estimate sweetness accurately from sensor values. The application loads a previously trained model, scaler, and feature configuration files to ensure consistency between training and deployment. It continuously fetches the latest sensor readings from the API, preprocesses the values, and passes them to the model for prediction. The predicted sweetness value is then converted into a meaningful quality grade such as A, B, C, D, or Below D based on predefined threshold ranges. The system also records the timestamp and displays the model used for prediction, making the process more transparent and informative. A Flask web framework is used to build the backend and serve the prediction results through both a web interface and a JSON API endpoint. This makes the solution lightweight, practical, and easy to integrate into real-time industrial or agricultural environments. The project demonstrates how machine learning can be combined with IoT-style sensor data and web technologies to create a smart prediction system. It reduces human effort, improves decision-making speed, and provides consistent quality assessment. Overall, this system offers a simple, scalable, and effective approach for sweetness estimation and grading, which can be further extended for smart farming, food quality monitoring, and automated classification applications.
 
INTRODUCTION
Sweetness is one of the most important quality indicators in food products, fruits, beverages, and several agricultural commodities. In many practical environments such as food processing units, storage facilities, fruit grading centers, and smart farming systems, the sweetness level of a product plays a major role in determining its quality, ripeness, consumer acceptance, and market value. Traditionally, sweetness is measured using manual testing methods, laboratory instruments, or human inspection, which may require more time, effort, and cost. In addition, manual evaluation may not always produce consistent results, especially when large quantities of samples need to be tested continuously. Due to these limitations, there is a growing need for an automated and intelligent system that can estimate sweetness quickly, accurately, and in real time. With the rapid advancement of machine learning, Internet of Things (IoT), and web technologies, it has become possible to build smart systems that can monitor sensor data and make predictions automatically. Machine learning is especially useful in such applications because it can learn hidden patterns from historical data and make predictions for new input values without requiring explicit rule-based programming. This project focuses on the development of a real-time sweetness prediction system using a trained machine learning model integrated into a Flask-based web application. The main purpose of the system is to predict the sweetness level of a sample using sensor values obtained from an external API and then classify the result into a quality grade for easy interpretation. The system is designed to act as a practical solution for real-time quality monitoring, where live data can be fetched, processed, and analyzed automatically without manual intervention.
In this project, a machine learning model is first trained using previously collected sensor data and sweetness-related output values. Once the best-performing model is selected, it is saved and later loaded into the deployment environment using Python libraries such as Joblib. Along with the trained model, the project also uses a scaler and feature column configuration to ensure that the same preprocessing steps used during training are applied during prediction. This is very important in machine learning deployment because the prediction input must follow the same structure and scale as the training data to maintain accuracy. The developed Flask application acts as the main platform that connects the trained machine learning model with live sensor data and the user interface. The application automatically fetches the latest sensor values from a given API endpoint using HTTP requests. These values are then extracted, cleaned, converted into numerical form, and arranged into a structured tabular format using Pandas DataFrame. If the selected model requires scaled input, the system applies the scaler before passing the data to the prediction model. The machine learning model then predicts the sweetness value of the given sample, and the result is rounded for better readability. To make the output more meaningful for users, the predicted sweetness value is further converted into a quality grade such as A, B, C, D, or Below D based on predefined threshold ranges. This grading mechanism makes the system easier to understand and more useful in real-world quality control applications.
The integration of Flask in this project provides an efficient way to deploy the prediction system as a web-based application. Flask is a lightweight and flexible Python web framework that allows developers to create backend logic, define routes, serve web pages, and return API responses with ease. In this system, Flask is used to create a home page for the user interface and an API endpoint for fetching live prediction results. When the user accesses the application, the Flask server loads the trained machine learning model and waits for prediction requests. Once the prediction route is triggered, the system fetches the most recent sensor data, processes it, and returns the final prediction output in JSON format. This design makes the application simple, responsive, and suitable for real-time deployment. The use of API-based sensor data collection also makes the system highly adaptable, as it can be connected to external IoT devices, cloud databases, or live monitoring platforms. Instead of manually entering values into the system, real-time sensor readings can be directly used to generate predictions, which increases efficiency and reduces the possibility of human error. This makes the proposed system a useful example of combining machine learning with real-time data communication and web deployment.
The importance of this project lies in its practical application and scalability. In modern agricultural and food technology sectors, there is an increasing demand for intelligent quality assessment systems that can support automation and smart decision-making. Sweetness is a valuable parameter in assessing the maturity and quality of fruits, juices, syrups, and other food products. A real-time sweetness prediction system can help farmers, quality inspectors, and industrial operators monitor product quality more effectively and take immediate actions when required. For example, such a system can be used to determine the best harvest time, classify produce into quality categories, monitor processing consistency, or support automated packaging and sorting systems. Since the system uses machine learning, it can also be improved over time by retraining the model with more data and selecting better-performing algorithms. In addition, the web-based nature of the system makes it accessible and easy to integrate with dashboards, mobile applications, or industrial control systems. The prediction output can also be stored for future analysis, trend monitoring, or report generation. Therefore, this project not only demonstrates a technical implementation of machine learning deployment but also addresses a real-world problem in a meaningful and efficient way.
Another important aspect of this project is its modular and maintainable design. The code is organized in a way that separates model loading, grade classification, data fetching, prediction logic, and route handling into different sections. This improves readability and makes the system easier to modify, test, and extend. The error handling included in the code also improves reliability by checking whether the API response contains valid data, whether the required fields are present, and whether the number of sensor values matches the expected model input features. If any issue occurs, the system returns a clear error message instead of crashing unexpectedly. This is a valuable feature in real-time applications where data inconsistency or network issues may occur. Overall, this project serves as a strong example of how machine learning models can be transformed from offline experiments into live intelligent applications using Python and Flask. It combines data science, real-time API communication, and web development into a single practical system. The proposed sweetness prediction system is simple in implementation, effective in performance, and highly useful for real-time quality assessment scenarios. It represents an important step toward smart monitoring and automation in agriculture and food technology domains, and it can be further expanded in the future with advanced models, richer user interfaces, database integration, and live dashboards for continuous monitoring and analysis.

OBJECTIVES
 To develop an intelligent sweetness prediction system using machine learning techniques.
The system should accurately estimate sweetness values based on input sensor data collected from external sources.
 To fetch real-time sensor data automatically from an API for live prediction.
This helps eliminate manual data entry and enables continuous monitoring of the sample quality.
 To integrate the trained machine learning model into a Flask-based web application.
The web platform should allow smooth prediction processing and user-friendly result display.
 To classify predicted sweetness values into quality grades such as A, B, C, D, and Below D.
This makes the prediction results easier to understand and useful for practical quality assessment.
 To create a scalable and efficient system for food or agricultural quality monitoring applications.

Block Diagram