2.1. Deep learning algorithms for enhancing decision-making capability in the area of business operations:
The stated section above elaborates the review of the present algorithms evolved in Deep-learning techniques in enriching the decision-making phenomena determined in Business operational-perspectives.
The challenge of language-ambiguity and the language complexity evolved in the emotions recognition prevailed in the narrative-documents to the most accuracy level. Hence there is a need to improvise the performance. Therefore, this can be accomplished by the process of deep-learning. Owing to this scenario, the paper describes this concept. The paper  demonstrates the specific-feature, which essentially needs the RNN-recurrent-neural-networks customization in accordance with the bidirectional-processing. Regularization layers of dropout and weighted-loss functionalities. The performance analysis of the paper is assessed in the six benchmark-data-sets. This methodology concluded that RNN-model and the transfer-learning overtake the existing machine-learning approaches.
Another to this concept, S2SCL-seq2seq based CNN LSTM approach, which is the one-step integration optimized decision-making process on the basis of deep-learning is implemented in the paper. This methodology incorporates the demand forecasting methods and the inventory-optimization process. This forecasting model performs the forecasting process and presents the architecture that incorporates the LSTM-network and the CNN-network. This proposed-framework  is capable of designing the dependency relations and the system-dynamics within it. Apart from the part of predicting the results, the methodology also quantifies the demand-uncertainty by the process of dynamic-distribution process. The model enables to bring out the optimized decisions relied on the service-capacity allocation in logistics. The inferences of the study revealed that S2SCL-model overtakes the other task bench-mark design models. Likewise, in the business's progress, another concept presented to uplift the organisation’s attention and the research persons to the several applications usage and the big-data benefits. Hence the paper  provides a detailed overview of the present trends, big-data pitfalls, big-data opportunities and the impact of the big-data technology to trigger the firms in creating proficient business approaches and the competitive background around the organizations.
Further to this, it also defines the Business-analytics applications and the characteristics of the data-sources. These sections of the review provide a clear picture of the larger-data-set manipulation and the better-management of the datasets. This is employed by the use of big-data tools and big-data techniques would create business value insights. Similarly, another study focusses on the fourteen counts of the micro-foundations analytics-enabled of dynamic-capabilities and also highlights the impact how the firms utilize the analytics in the Operational research management and the improvisation Also, some of the business-aspects and the technical-aspects in OR-studies refer to leverage the dynamic-business analytic phase. In order to fulfil the gap, the paper brings out the firm's capabilities perspective and also constructs the eight operational studies that relied on upon large-scale organisations. Every organisation has a significant impact on the analytics phase and the implementation phase.
Another article brings out the strategy to compare the various deep-learning techniques for data-processing with distinct hidden-layers and the neurons count.  The comparative analysis of the study presents that Deep-learning can be constructed through the un-supervised and super-vised training approaches. Such type of strategy improvises the business plan for the organisations. Such kind of approach again focussed on another study. The goals of the article are stated as follows, such as the deep-learning review for the BA-business analytics process from the operational-perspective. The second goal provides the motivation of the reason why the business-analytics practitioners and the business-researchers use the DNN and gives out the glance of the capable use-case, requirement analysis along with the benefits of the same. On the third objective, the additional operational value, along with the real-time data, originated from the enterprise's undertakings. These cases illustrate the operational performance enhancements upon the existing machine-learning techniques. Further to this, the implications and the business-guidelines involved in OR is presented for the BA-business-analytics capabilities according to Deep-learning methods. The fifth objective focusses on the summation of the experimental analysis. This depicted that the out of the box design architectures are sub-optimal and insisted on the customized-architecture value in modelling the new deep embedded-networks.
The section of the article adds to the utilization of DL-applications for the effective decision-making process of the firms. It brings out the DL-algorithms tutorial overview, describes the DLADM-processes with the tasks of image-recognition, and the sentimental-analysis relied on Zalando data-sets. In the final section, it is also discussed the DLA-DM challenges and DLADm-promises. It has been a major challenge to compromise the practitioners of health-care to employ the Business-analytics technologies in the transformation of the health-care industry and to gain value. Hence as the measure, one study  Moves forward to bring out the research-questions related to Big-data present-works. The first step is determining the BA-application's essentialities, where the health-care organization would attain in taking the successful-strong decisions. The second discussion the list of all the organizational-abilities which provokes the health-care firms to deliver the decision knowledge, usage of the BA-Business-analytics systems for the decision-makers and in the decision-making of the stake-holders. The next part of the article reviews the present research studies relying on BA-business value exploration. The experimental-analysis of the research involved examining BA-theoretical background, the efficiency of the decision-making process for modelling the proficient research-framework. This model provides an overview of the period when to start the Ai-artificial-intelligence phase in the business so that the researchers or the businessman can arrange the organization in learning the BA-phases, exploitation of the old strategies, and enhance the new-potentials. The perceptions of the study rely on the analysing phases of the business-baselines, researching phase, and in manipulation of the real world use-cases and illustration of the leaders, scientist’s collaborations status.
2.2. Effective and accurate Big-data processing based operational parameters:
The following section of the paper illustrates the review of the existing studies implementing the Big-Data technologies efficiently to contribute to the impact on operational parameters.
The article also employed a well organised evaluation, and the analysis of the big-Data is implemented in the companies. This study  presents the BDA-architecture, which comprises of the 6 components such as the data-generation, acquisition of the data, storage of the data and the advanced techniques of data-analytics, visualization techniques of the data and the value creation of the decision making processes. The characteristics of Big-Data analytics is demonstrated in the paper, such as the velocity of data, volume of the data, variety of the data, valence factor, veracity-factor, variability factor and data-value is elaborated. The article also reviews the complications of the study such as the BDA-concepts, characteristics of BDA, processing paradigms of Big-Data-analytics, state of the art model BDA-framework evolved in the decision-making processes to a vision of the BDA-values. Also, it states the present Big-Data analytics challenges and the plans associated with them.
Some of the Big-Data Analytics has an impact on other streams as well. Hence illustrating such type of approaches, this study,  provides an overview of the resources of the remote-sensing data, present progress of the BDA-remote sensing technologies, processing of the remote-sensing data and the management of data. Therefore, such a model of a five-layer of fifteen-level framework model referred to as FLFL, this type of satellite-remote-sensing management of data design is illustrated. Hence, the FLFL-four layered twelve-level structure of data-management and the remote-sensing BDA agriculture application is presented for the factor of precision. In this model, the sensors relied on the higher resolution satellites, aircraft, aerial-vehicle of unmanned types and the ground basis structures. This features the forecasts of the remote-sensing BDM future organisation and the local-regional application, and the farm-scale applications.
One of the Big-Data concept, such as the application in the diagnostic platform of the images, is implemented in the article. The paper evolves the obstetric-diagnostic stream of images on the basis of cloud-computing technology. At the first level, the medical-imaging process is created by integrating the cloud-computing, distributed file-systems and caching techniques.  In the second stage, the contrast improvises ultra-sound technology yields the most image accuracy in the factors of structure, developmental-abnormalities, location-factor of the placenta. At the concluded inferences, the imaging diagnostic-platform efficiency undergoes some verification processes for efficiency. The results of the paper depict that the platform-framework acquires faster-data processing and ease of use. These facilities significantly decrease the medical-equipment costs and enhance the efficiency of the framework. In this paper,  the study presents the literature review related to the utilization of IoT-internet of Things framework and the Deep-learning methods to enhance the smart-cities. The various infrastructures of computing environment utilized for the IoT-Internet of things BDA is illustrated. This IoT-BDA technique includes the cloud-computing, fog-computing and edge-computing platform. The popular DL-Deep-learning designs are described in the survey, and the present research studies are evolved to design the smart-devices and the smart-applications. The IoT-technology is defined and brings out the computing technology infrastructure utilized by the IoT BDA processes. In the conclusion part, the critical DL-Deep-learning challenges and the open complications are highlighted while modelling the IoT-Smart cities apps.
Another chapter implements the novel big-data solution pipeline for sensor-data storage and in the data-processing. This proposed framework  processes the sensor data by utilizing the Apace-Flume for the effective data transformation, and the IoT-data-collection originated from the cloud-computing server. This Data transformation is transmitted to the Hadoop-distributed storage file-system. And also, this Apache-storm employed to process the factual time information. In the next step, the researchers proposed the utilization of a hybrid-prediction model of DBSCAN-Density-based spatial-clustering of apps with noise to eliminate the outliers of sensor-data. This also yields better fault-detection accuracy rates by implementing the classification processes of the SVM-support-vector-machine algorithm.
Another detailed paper to provide the overall review of the present research studies relied on deep-learning models for the Big-Data feature-learning. At the first stage, the four kinds of DL-Deep-learning models, such as the Deep-belief network, stacked auto encoder technique, RNN-recurrent neural-networks and the convolutional-neural networks, is explained.  These models is utilized in the feature-learning of Big-Data. The following section again presents the Deep-learning models overview in accordance with the 4V's design model, such as the larger scale DL-data handling the larger amount of data, heterogeneous model computation designs, multi-modal Dl-models, incremental DL-models for the real entity information and the Dl-model with good reliability for lower qualitative information. In the final section, the current DL-Big-data challenges and depicts the trends of the BDA-process. Owing to providing solutions to security attacks in the IoT environment, another concept is discussed in the following paper. This works as the solution to determine the new-threats of low-false positive rate and high percentage of detection. Also, it defines the contextual attacks and the collective-attacks in security.
In this paper  the language-processing concepts, distributed-deep-learning concepts, Big-data concepts, flow-analysis of anomaly-identification concepts and the contextual-analysis concepts are integrated to generate the model. Further to this, the framework defines the network-abstract behaviour obtained by the millions count of packets in the context. This paper also evaluated the work in real time to the destination point involving the contextual anomalies and the collective anomalies.
Likewise, the paper depicts the big-Data evolution in the IIoT environment and also elaborates the brief associated technologies survey, including various algorithms, case-studies related, and the framework associated with the study.  And the brief taxonomy is elaborated in the key-concepts classification. These related frameworks the case-studies are discussed in the paper. Similarly, concentrating on future plans, the brief future-opportunities discussion, technologies concepts and the research problems are also outlined in the paper. The present BDA-systems presents the data-engineering frameworks, data-engineering preparation and data-analysis. But even some attempts are necessary to change the present BDA-methodologies in meeting the IIoT-systems requirements.
2.3. Deep learning based big data applications
The approach of Deep-learning methodology has an immense role in the manipulation of Big-Data analysis of Big-Data Access in various platforms. Hence this section focusses on some of the Big-data manipulation applications after employing the Deep-learning techniques.
Hence in taking consideration of these points, one of the studies illustrated the big-Data analytics significance.
And the importance of computational-intelligence strategies is elaborated. These concepts were employed in the generated data from the embedded-personalized devices and the distributed data-processing techniques. It also lists out the summation survey  of the computational-intelligence approached for Big-data analytics and efficient big-data processing. This study also enumerates the Data-modelling process, which brings out the HSTSM-Hierarchical-spatial Temporal-State-machine approach and new generative modelling design of biologically-universal technique. Another application of the Deep-learning system in handling the Big-Data is illustrated in the study. This study enumerates the step-wise processes of how the proto-type-Deep learning application employed on GPU-clusters and the CPU-clusters. The assisting guide tools are Python technology and Redis-technology. This research  exhibits the good understandability of the readers in the construction basis of the GPU-application and the distributed-performance level of the system within less number of hours. This system does not have a dependency on the application of Deep-learning and Deep-learning framework. The lower-level of construction-blocks is utilized so that the model can be managed to any parallel-algorithm and the reader rely on Big-Data-prototype. At the conclusion step, further discussions also explained how the model could be moved out from the prototype-level to the full-fledged production application.
The Impact of Big-Data analytics for handling the complex-type of data is illustrated in the study stated as following. This work  exhibits the influence of big-Data analytics to be employed as the effective-process to handle and rectify complex and un-structured information. This is accomplished by utilizing the technologies including spark, Map-Reduce method and the Hadoop-technology. Further in this study, this paper also describes the Big-Data challenges in accordance with the literature section comprising of Six-v’s such as volume-data, veracity-concept, velocity-concept, variability, value-concepts and variety-concepts. The case-studies of Big-Data relying upon several techniques such as text-data analysis, voice-data, video information and network-analytics data is investigated in work. It is concluded that the analytics of Big-data would pose out the positive implications in the fields such as the health-care field, Business-field, banking-sector, marketing-sector, politics-sector and the military-sector. Focussing on the other concepts like IoT environment involving the Deep-learning techniques is illustrated in the study. The paper aims to provide the clarification of multi-disciplinary strategy on the basis of IoT-framework in the deep-learning approach. In this IoT-approach Framework, the collaboration of the various expertise-persons, data-scientists, smart-city infrastructure, system-driving concepts. It is also depicted how the individuals of multi-disciplinary sides perform the interactions, design processes and perform the implementation of business-oriented applications of Deep-learning concepts. The Business-analytics operational system reveals how the IoT-analytics systems would be efficient enough to depict the inferences of analytics in the graphical representations as well. Hence the study also figures out some of the examples of applications employed in the multi-disciplinary-process and in the assessment of efficiency.
 The applications of Big-Data Analytics is employed in the improvisations in the application-intelligence concepts involved in the field of transportation. Hence to illustrate this process, the study has been stated below. The utilization of Big-data algorithms increases day by data and acquire the academic-attentions and the industrial-field attention level in ITS-Intelligent-Transportation system. This algorithm of Big-Data involved in ITS- applicable to many applications. Still, they do not have the boundary to the phenomena of signal-recognitions, predictions of traffic-flow, planning of travel-time, route-planning of travelling, object-detection methods and in the safety measures of the road and the vehicle premises. Hence as the formulations to this concept, the study is evaluated. This work has the objective to bring out the review studies of the ITS-applications and the review studies of the Big-Data-models utilized in ITS-context. As a result of inferences, the study would provide the depth-insights of the Big-data algorithms within the real-time applications involved in the ITS-method. In this study, almost five-hundred and eighty-six papers have been reviewed in the period of 1997 to the year of 2019.
Similarly, another study stating the Deep-learning approach in handling the forecasting of time-series involved in Big-Data is illustrated. The deep-forward neural-network utilized in the Apache-spark platform in the distributed-computing forum.  The evolution of H2o analysis of Big-Data do not permit the multi-step regression arrangement, and the methodology utilized for the arbitrary length horizons is implemented. In this prediction, the future values count is predicted. The real world dataset results comprised of the electricity-consumption possessing the frequency rate of sampling from the year 2007 and ending in the year 2016 were inference. The runtimes-parameters and the accuracy-factor versus the resources of the computing platform, along with the dataset size, have been assessed. In the conclusion part, the scalability-factor of the proposed-framework made in comparison with the other existing techniques, which depicted the sufficient methodology in processing the time-series of Big-data.
2.4 Data dimensionality and other big data issues solving using deep learning and its feature extraction techniques:
The techniques stated above also plays a vital role in the rectification of Big-data complications in the analysis phase by implementing the deep-learning technologies.
The study elaborates and integrates the researchers of machine-intelligence fields and the cyber-security fields to enhance the anticipating-missions, prohibiting-missions, preventing measures, preparation-measures, the response-factors to the several complications of cyber-security and the related challenges associated with it.  The broad discussions of the topics illustrated in the book provide the readers in their multiple views relying on the various machine-intelligence-disciplines and cyber-security applications. The machine-intelligence-concepts and the analytics of Big-Data for cyber-security-apps compose of the various state of art implications received feedback from the practitioners and the machine-intelligence-scientists and cyber-security-field.
In the scenarios to provide the dimensional reduction basis solutions, it would yield out the spectral data loss, which has an impact on the performance level of the classification-techniques. Hence as the remedy to refer to the complication, the paper  proposes a framework known as SAS_DBN-spectral-Adaptive Segmented DBM-technique for the HIS-technique. This approach would exploit the features from the segmentation process of the genuine spectral-brands to the smaller spectral-brand groups. These groups were processed individually utilizing the local-DBN. The experimental assessment of the study relied on the HSI-data-set of standards with various context, and different-resolutions promotes the framework efficiency level. The resultant outcomes made in comparison to the different present HIS-techniques of classification. Likewise, another procedure is developed for the predictions of JRT-job-remaining time. The three sections of the JRT-predictions are collecting raw-data, candidate design of the dataset and the modelling process of prediction. In this first section, the production information of historical concepts were gathered by the IoT-deployment technique. After the section, the dataset of the candidate is formulated for the predictions of JRT. This formalisation is carried out to obtain the JRT-contributory factors in the predictions. This study is depicted as the very first-Deep-learning model in the JRT-predictions dynamically in the production-phase. These methodologies of production employed in the larger scale job-shop, which is equipped with the forty-four machine-tools and it generated the thirteen part-types. The results from the experiments revealed that the S-SAE-design acquire the high accuracy-rate than the past regression-model, back propagation-network and the multi-layer network design involved in the predictions of the JRT-process. A similar analysis of Big-Data is described in another study illustrated below. The HIS-Hyperspectral-imaging technique is employed in the prediction of TVB-N-total-volatile primary nitrogen content prevalent in white-shrimp. For this purpose, SPA-successive-projections algorithms and the SAE-deep learning basis stacked-auto-encoders algorithms were used integrated with the spectral extraction of the features. For the prediction-process, LS-SVM-least-squares support-vector-machine, PLSR-methods and the multiple type linear-regression technique is utilized. The study results revealed that the prediction models of SAE-design acquired a better performance level than the other prediction-models of SPA-basis.
2.5. Big data processing using deep learning for decision making from predictions:
Some of the applications of Deep-learning approaches contribute to the primitives of decision-making mechanism in the organisational sectors. Such studies related to this is elaborated in the section. The paper illustrates the AutoML-automated-machine-learning technique in the predictions of risk and the evaluation of behaviour. These predictions are utilized in the decision-making processes and in the motion-trajectory process of planning in the AV-autonomous-vehicles. This methodology  attains the higher efficient results in the risk predictions of behaviour-basis. This revealed the predictive percentage of 91.70 in the total accuracy rate in the 4 risk-levels and 95% rate in the safe-risk distinction. The model incorporates the 2 major components, such as the non-supervised identification of risks, Feature-learning techniques and auto-tuning process of the model implemented by the Bayesian –optimization process.
Some of the studies illustrating the deep-learning application concepts are defined in the study. The paper has the objective to yield several Big-data applications indulging the deep-learning strategies and the eliminating the multi-criteria strategy for rectifying the Big-data analytics complications.  Also, additionally, the various fields involving the information-technology, Business-fields, agriculture-domains, Computer-science employs Deep-learning techniques and multi-criteria basis decision-making conflicts.. The study presents the Big-data service-architecture, which encapsulates the data gathering and the storage-process. For the information gathering process, the technical-processing methodology is also implemented.  The following process is that the paper also discusses the processing techniques of Big-Data and the big-data analysis phase according to the various service requirements. These service-requirements provides the valuable-information to the service customers. Also, the cloud-computing service operational system also introduced on the basis of Big-data. This system would facilitate in providing the higher efficient solutions in handling the larger scale data processing methods, data-analysis, and in the larger size of data.
Another concept of sentimental-analysis is discussed in the following study. The sentimental-analysis approach is illustrated in the study to adopt the fast-Text along with the variants of RNN-Recurrent-neural-networks. This approach is employed to efficiently represent Textual-Data. After the process, the classification technique representations is performed in the study. The primary goal of the study is to improvise the RNN-recurrent-neural-network performance by the classification technique accuracy-factors and in the management of larger-scale information. The results of the experiment reveal that the proposed-framework is capable to improvise the three-model performances. The present approach would yield the facilities to the big-data practitioners and the big-data researchers who are in need to gather the data, manage the data in data-analysis and who proceed with the visualization of the different information sources in real-time entity. To concentrate on the challenge study related to the AI-based operational systems is discussed in the study following. The study presents the identification of challenges associated with the usage of AI-based operational systems and challenges associated with the influences of AI-based systems.  These systems facilitate in the purpose of the decision-making process, and it also it provides the research-preposition set for IS-information-system researchers. The study yields the view of AI-based history data by giving the related studies in IIJM-International-journal of Information-Management. Hence in the advance studies of the researches of AI-usage in the Big-Data era, the study provides the 12 prepositions of the research in accordance to the terms of theoretical development factors and the conceptual-framework factors, AI-system-human interaction concepts and in the implementation of AI-systems. Similarly, another study involves the optimization of the Deep-learning algorithm parameters to predict the infections of the diseases is employed in the study.  This study considers the social-media information as Big-Data. Further performance analysis of the DNN-model and the LSTM-learning model is exposed to comparison with the ARIMA-autoregressive-integrated moving-storage model. This performance is analysed in the prediction process of the infectious-diseases in 1 week. The resultant data in the study exhibits the efficient performance of the DNN-model and LSTM-model rather than the ARIMA-model. The integrated models showed a performance rate of 24.0% and 19.0% in the prediction of chicken-pox disease of the top-10-DNN-model and the LSTM-learning-model. Likewise, another framework involved in the review analysis is this paper. This paper has transformed the negative-user opinions and positive-user opinions in the quantitative type scores. In this paper, sentimental-analysis is performed in the assessment of Amazon-online reviews.  The FRDF-Fake-Review Detection-framework have determined and eliminated the fake-reviews by utilizing the NLP-Natural-language-processing method. The FRDF-model is evaluated on the product-reviews categorized from the higher-technology industries. The Brands of the products are subjected to the rating process in accordance with the sentiments of consumers. The inferences of the study described that the managers of the business and the online-consumers utilize the tool and take it as an effective decision-making process.
2.6. Big data processing using deep learning in larger datasets:
A massive amount of data is also employed in the processing of Big-data through deep-learning techniques. The studies related to the work is described below.
It is necessary to focus on the aforesaid-context because the automation of the data-processing technique is not always preferable since it costs more in the data-analytics phase.  Hence this study is implemented in highlighting the engineering-data specificities and the complications of data-processing techniques originated from the manufacturing-firms. Therefore the effective approach in employing the artificial intelligence process to yield the efficient methodologies and the proficient tools to overcome the above mentioned issues of the data. Hence in this study, a special focus is provided to bring out the literature-review of the present applications, which could outstand to promote the enhancement to the machine-learning methods and the deep-learning methods., As the inferences to the study, the outcomes of the studies would pave the way to the enhancement to the open source dataset comprising of the two-thousand CAD-learning models. Similarly to this study above, again, a detailed survey on the state of art methods of deep-learning algorithms, Big-data studies and the IoT-system studies.  Additionally, an analysis of the comparisons and the associations between the deep-learning technologies, IoT-system studies and the Big-Data learning-technologies are also focussed. There has been established the thematic-taxonomy from the inferences of the comparison-analysis phases. The integration of deep-learning technologies for IoT-system security utilizing the big-data processes faces some of the challenges. These challenges were also discussed in the conclusion part of the paper. Future works have strived the paths to the researches relying on the securing primitives of IoT-systems.
In the field of the Latest Trends of technologies, the predominate features of Big-data are the process of Heterogeneity, where the heterogeneous resulted in the data-integration issues and the analytics phase problems of Big-Data. Hence to remedy to the issue, the study  presents the heterogeneous data-processing methodologies, phases of Big-data analytics, tools of Big-data-analytics, methods of Machine-learning technology and the traditional- Data-mining techniques were assessed. The beneficiary facts of the Big-Data analytics phases, HPC-higher-performance computing phases and the Heterogeneous-computing phases are also discussed. The challenges pertaining to handling the Big-Data and the heterogeneous-data were also described.
One of the application in handling the huge-data-set involving the Deep-learning technologies is employed in the study. In this study, an ABC-data-set is presented wherein the CAD-models collection is carried out for the geometric-deep-learning process researches and the Deep-learning applications. Every model is depicted as the gathering of parameterized-surfaces and the parameterized-curves explicitly.  This data would provide the truth for the several segmentation of the patches, different-quantities, feature-detection methods and the reconstructions of the shape. Hence in the overall inferences to the study, the larger dataset scale is performed for the prediction process and in the evaluation of the performance towards the other existing methods of estimation process. Similarly in the analysis phase of handling the massive-data amount necessitates the enhanced techniques for proficient review or estimating the future plans to be executed, in providing the higher-precision methods and the bringing out the better decision-making approaches. Hence as the objective point to the statement above, a little study  has been performed in the uncertainty condition for the analytics of Big-data processing and also in the AI-techniques applicable to the Big-data data-sets. Since the amount of Data varies to a massive level, the data variation and data-speed increases, leading to the uncertainty condition. This would turn lead to a lack of confidence-level in Big-data analytics and the decision-making process. Further to this, the article provides a review of the past big-data analytics studies and bring out the challenges of the studies. Also, it states the future plans in uncertainty recognition and in the mitigation of uncertainty in the respective-domain.
Owing to this concept as well, the novel approach in considering the larger scale, real time, faster traffic-predictions and bring out the emerging technologies such as the Big-data technology, Deep-learning technology, GPU-Graphical-Processing-Units and the in-memory computing-techniques. In this methodology, Deep-networks were trained upon the eleven years of data (Caltrans-department), and the large size of the dataset was utilized in the deep-learning study works. And in a result, various input-attributes combinations, with the other several Deep-learning configurations of the network, is analysed for the prediction techniques and for the training processes. The pre-trained design model utilization is brought out for real time prediction-processes. The paper also elaborates the novel-models of deep-learning methods, Implementation of the algorithms, methodology of the analytics and the Smart-cities software-tools, computing efficient performance and the convergence rate of the Big-data-analytics.