This study proposes a hybridized deep neural network-based framework for sentiment analysis, where the Dispersive Flies Optimization (DFO) algorithm is modified by adjusting its neighbor counterpart. The modified algorithm, called Neighbour Adjusted Dispersive Flies Optimization (NADO), is used to optimize the feature space with the aid of sentiment information extracted using a specially developed SentiWordNet lexicon-linked fitness function. This modification helps avoid local optimal solutions and supports the optimization process in approaching the global optimal solution more effectively. To handle textual features efficiently, pre-trained embedding techniques are used to represent them mathematically. The hybridized deep neural network, composed of a Convolutional Neural Network (CNN) and Long Short Term Memory (LSTM), is then given the embedded features. CNNs construct hierarchical representations to store locally implanted information, while LSTMs attempt to recollect relevant prior data for opinion categorization. This hybridization allows the system to take advantage of both component networks. The deep neural network system ultimately delivers the desired sentiment category. The proposed framework is tested and compared with various state-of-the-art methods using multiple performance indicators. The framework achieves the best performance compared to the baselines with accuracy rates of 89.0%, 81.9%, 67.9%, 64.6%, 83.2%, 79.8%, and 91.3% on Amazon, ETSY, Big Basket, Facebook, Finance, Twitter, and Wine datasets, respectively. The study highlights the effectiveness of the hybridized deep neural network framework in sentiment analysis. The framework leverages the strengths of CNNs and LSTMs, and the modified NADO algorithm enhances the optimization process. The study also discusses the importance of sentiment analysis in the context of customer feedback and the role of neural networks in modern machine learning applications.This study proposes a hybridized deep neural network-based framework for sentiment analysis, where the Dispersive Flies Optimization (DFO) algorithm is modified by adjusting its neighbor counterpart. The modified algorithm, called Neighbour Adjusted Dispersive Flies Optimization (NADO), is used to optimize the feature space with the aid of sentiment information extracted using a specially developed SentiWordNet lexicon-linked fitness function. This modification helps avoid local optimal solutions and supports the optimization process in approaching the global optimal solution more effectively. To handle textual features efficiently, pre-trained embedding techniques are used to represent them mathematically. The hybridized deep neural network, composed of a Convolutional Neural Network (CNN) and Long Short Term Memory (LSTM), is then given the embedded features. CNNs construct hierarchical representations to store locally implanted information, while LSTMs attempt to recollect relevant prior data for opinion categorization. This hybridization allows the system to take advantage of both component networks. The deep neural network system ultimately delivers the desired sentiment category. The proposed framework is tested and compared with various state-of-the-art methods using multiple performance indicators. The framework achieves the best performance compared to the baselines with accuracy rates of 89.0%, 81.9%, 67.9%, 64.6%, 83.2%, 79.8%, and 91.3% on Amazon, ETSY, Big Basket, Facebook, Finance, Twitter, and Wine datasets, respectively. The study highlights the effectiveness of the hybridized deep neural network framework in sentiment analysis. The framework leverages the strengths of CNNs and LSTMs, and the modified NADO algorithm enhances the optimization process. The study also discusses the importance of sentiment analysis in the context of customer feedback and the role of neural networks in modern machine learning applications.