This document proposes a recommendation system for e-commerce called SBT-Rec that is based on structural balance theory. It aims to address challenges with sparse user rating data where traditional collaborative filtering may not find similar users or items. SBT-Rec first identifies a target user's "enemies" or opposite preferences, then determines "possible friends" according to the rule that "an enemy of my enemy is my friend". It recommends items preferred by these possible friends. It also identifies "possibly similar items" for a target user's preferred items using the same rule. The document outlines the SBT-Rec algorithm and describes its architecture which collects data, performs recommendation, and provides a user interface.