Big data analytics and processing requires complex architectures and sophisticated techniques for extracting useful information from the accumulated information. Visualizing the extracted data for real-time solutions is demanding in accordance to the semantics and the classification employed by the processing models. This article introduces fuzzy-optimized data management (FDM) technique for classifying and improving coalition of accumulated information based semantics and constraints. The dependency of the information is classified on the basis of the relationships modeled between the data based on the attributes. Different from the conventional fuzzy-decision systems, this technique segregates the considered attributes based on similarity index boundaries. This helps to process complex data in a controlled time and also supports large data processing. The performance of the proposed FDM is analyzed using a real-time weather forecast dataset consisting of sensor data (observed) and image data (captured). With this dataset, the functions of FDM such as input semantics analytics and classification based on similarity are performed. The metrics classification and processing time and similarity index are analyzed for the varying data sizes, classification instances, and dataset records.

Internet of Things (IoT) has gained a significant importance due to its flexibility in integrating communication technologies and smart devices for ease of service provisioning. IoT services rely on heterogeneous cloud network for serving user demands ubiquitously. Service data management is a complex task in this heterogeneous environment due to random access and service compositions. In this manuscript, a machine learning aided information management scheme is proposed for handling data to ensure uninterrupted user request service. The neural learning process gains control over service attributes and data response to abruptly assign resource to the incoming requests in the data plane. The learning process operates in the data plane where requests and responses for service are instantaneous. This facilitates smoothing of learning process to decide upon the possible resources and more precise service delivery without duplication.

Magnetic Resonance Imaging (MRI) inputs are most noticeable in diagnosing brain tumors via computer and manual clinical understanding. Multi-level detection and classification of the images utilizing computer-aided processing depend on labels and annotations. Though the two processes are dynamic and time-consuming, without which the precise accuracy is less assured. For augmenting the accuracy in processing un-labeled or annotation-less images, this article introduces Absolute Classification-Detection Model (AC-DM). This model uses a conventional neural network for training the morphological variations proficient of achieving label-less classification and tumor detection. The traditional neural network trains the images based on differential lattice morphology for classification and detection. In this process, training for the lattices and their corresponding gradients is validated to improve the precision of the regional analysis. This helps to retain the precision of identifying tumors. The variations are recognized for their lattice mapping in the detected boundaries of the input image. The detected boundaries help to map accurate lattices for adapting morphological transforms. Thus, the partial and complex processing in detecting tumors is restrained in the suggested model, adapting to the classification. The efficiency of the suggested model is verified utilizing accuracy, precision, sensitivity, and classification time.