Design Of Prediction System For Anticipating The Consumer's Purchase Intention Of Durable Goods

---

Abstract

This study captures a number of shoppers, online searches, and selection state in order to anticipate their purchase intent for durables. Amazon was chosen as an e-commerce platform to collect real-time search and review data from the client. In general, Amazon predicts the customer's purchase intent and promotes the goods in a variety of ways on its website. The suggestion approach was developed with the goal of making the program simple and user-friendly in the e-commerce industry, and research in this subject is still ongoing. Amazon's recommendation algorithms, which have been successful since 2003, employ item-based collaborative filtering. We used the Amazon product database for this study and projection, which contains over 1,500 user reviews for various Amazon items, including the Fire TV Stick, Kindle, and more. The dataset contains basic product characteristics, rating, review text, and more for each product. A powerful e-commerce platform was created for customers by developing a prediction model with attribute level decision support. To build the prediction model, the social perception score of brands and the polarity of feedback are calculated using social network mining and sentiment analysis, respectively. In order to forecast the relevant product attributes for each attribute, a suitable regression analysis and appropriate cases were then constructed for each attribute. In order to use the SVM Algorithm to execute and forecast the model more correctly, we incorporated some additional potent factors, such seasonality and polarity.

---

Keywords

Support Vector Machine (SVM); Purchase intent (PI); Amazon Web Services (AWS); Machine Learning (ML); Prediction Model (PM)