The research proposes a data-centric intelligent design in signal enhancement, noise filtering and predictive modeling to overcome the shortcomings of the traditional model-centric research designs. The model combines both the complex signal processing algorithms the Wavelet Transform and Kalman Filter, with the deep learning algorithms Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks. The main goal is to enhance quality of data before being subject to predictive analysis to improve overall performance of the system. The results of the experiment show that significant improvements were made, where the signal-to-noise ratio (SNR) had increased on the range of 12 dB to 32 dB and mean squared error (MSE) had decreased on the range of 0.020 - 0.006. The proposed framework had a high prediction accuracy of 97.8% that was higher when compared to traditional machine learning models and other hybrid methods. As well, the success rate of feature extraction rose up to 96.2% which proves the effective integration of deep learning. The comparative analysis with similar work further confirms that the proposed system is better, in terms of noise reduction, prediction accuracy, and robustness compared to similar work. The trade-off has slightly increased the computational cost, but produced a more reliable and scalable solution. The result that has been achieved in this study is the significance of data centric approach in the development of the intelligent systems within the real world situations like the healthcare, IoT and smart environments.