Smart Home Energy Management Systems (HEMS) constitute a vital necessity for optimizing electricity usage and saving energy in smart grids. However, these systems rely on dynamic factors that are stochastic and difficult to predict, such as the load consumption and electricity prices. Therefore, constructing an efficient control system for residential buildings requires an accurate prediction process of the associated parameters. This paper proposes an integrated predictive control system that consists of both predictive model and Demand Response (DR) scheme to predict and control the daily electricity usage in the residential sector. First, a Long Short-Term Memory-based (LSTM) optimized predictive model is implemented for predicting both the hourly load consumption and electricity price for a typical smart home. Then, the predicted data are transmitted to a DR fuzzy logic-based controller that can optimally schedule the home appliances usage. In comparison with the state-of-the-art prediction techniques for the residential load consumption and electricity price, the proposed LSTM predictive model outperforms Linear Regression (LR), Decision Tree (DT), Support Vector Regression (SVR), and Ensembled Boosted Trees (EBT). Moreover, the proposed DR-FIS controller has shown good results in terms of reducing the electricity cost by selecting the optimal time schedule.