Page 169 - Emerging Trends and Innovations in Web-Based Applications and Technologies

International Journal of Trend in Scientific Research and Development (IJTSRD) @ www.ijtsrd.com eISSN: 2456-6470
marketplace, setting a benchmark for efficiency, trust, and each tagged with their respective categories, condition
environmental responsibility. ratings, and historical price data. Table 1 outlines the
III. PROPOSED WORK number of product categories in the dataset.
In this phase, we define the process to revolutionize the Table 1. Number of product categories in the dataset
second-hand marketplace through Quick Mart's innovative Sr. No Category
solutions. The approach is based on optimizing the buying 1 Electronics
and selling process for second-hand goods using advanced 2 Furniture
technologies. The proposed framework for Quick Mart is 3 Clothing
presented below, where different data sets (training and 4 Books
testing) were used for prediction and classification tasks. In 5 Kitchen Appliances
the first step, products were categorized based on various
6 Sports Equipment
conditions (new, like-new, lightly used, etc.), and were
7 Tools and Gadgets
processed into a structured format for feature extraction.
8 Jewelry
After this, smart algorithms were applied for product
9 Musical Instruments
classification and price prediction.
10 Collectibles
The process is divided into four key sub-sections: data 11 Miscellaneous
collection, data pre-processing, classifier description, and
performance assessment. Each of these is described in Table 2. Number of images in model evaluation
Number of images Folder directory
further detail below:
4736 Training
Data Collection 1184 Testing
For this work, data was collected from Quick Mart's platform, 1184 Validating
which includes a variety of second-hand products across
categories such as electronics, furniture, clothing, and books. Validation Set: This set of images will be used during
The data was obtained by combining datasets from public training to adjust all the model parameters.
repositories and Quick Mart's own database. The dataset Testing Set: This set of images will not be involved until the
consists of thousands of images of second-hand products, final performance of the model is assessed.

Fig. 2. Sample images of second-hand products in the dataset.
Data Pre-processing
Data pre-processing plays a crucial role in any machine learning model. In this phase, missing values and redundant data are
handled. Data augmentation is applied to expand the dataset. The steps include:
 Loading the Data: The dataset is loaded, and the training and testing data are separated into arrays.
 Shuffling and Splitting the Data: The data is shuffled and then split into training, validation, and testing sets in an 80:20
ratio.
 Encoding Labels: Since the labels are textual, they are transformed into numerical values using Label Encoder.
 Converting Labels to Categorical Form: The labels are further converted into categorical form to enhance model training
performance.
Feature Extraction
Feature extraction plays a vital role in the success of any model. In the context of Quick Mart, feature extraction involves
identifying and isolating meaningful patterns or characteristics from the product images. These extracted features help to
predict product categories and their price range.

The feature extraction methods used include:
 Texture Analysis: Texture features, such as entropy and homogeneity, are used to evaluate product conditions and
identify defects.

IJTSRD | Special Issue on Emerging Trends and Innovations in Web-Based Applications and Technologies Page 159