This research investigates the viability of Long Short-Term Memory (LSTM) systems, a subtype of Recurrent Neural Networks (RNNs), for picture captioning. Leveraging the MS COCO dataset, the study compares the execution of LSTM-based RNNs with Vanilla RNN, Gated Recurrent Unit (GRU), consideration components, and transformer-based models. Experimental comes about to illustrate that the LSTM-based RNN shows competitive execution, accomplishing a BLEU-4 score of 0.72, a METEOR score of 0.68, and a CIDEr score of 2.1. The comparative investigation uncovers its prevalence over Vanilla RNN and GRU, highlighting its capability to capture long-range conditions inside successive picture information. Moreover, the study investigates the effect of consideration instruments and transformer designs, exhibiting their potential improvements in the context-aware caption era. The transformer-based show outflanks all other models, accomplishing a BLEU-4 score of 0.78, a METEOR score of 0.72, and a CIDEr score of 2.5. The findings give important bits of knowledge toward the creating scene of picture captioning strategy, which makes LSTM-based RNNs solid and productive approaches for capturing worldly groupings in visual substance. In achieving these, the study provides a framework for future developments in hybrid models and manufacturing processes that push boundaries of smart image perception and understanding
Recommendations can be used in two ways, one for the mass audience and second for personalizing, in the mass audience like weather recommendation, stock recommendation, so this type of recommendation is based on past generic data. However, personalized recommendations are based on unique dataset of users, and at the time of changing location and demographic condition changing frequently, it becomes more complex. This type of recommendation entirely depends upon the dynamic profile of the user, and this dynamic behavior can alter by the third person for fake and biased suggestions. In this paper, we identify the way of altering the recommendation system (RS), and at the same time, we propose a framework TRS for trusted RS for dynamic dataset based. This paper defines the process for fake profile classification, detection, and prevention from profile injection. In this framework we propose two phases, in first phase online social network (OSN) fake profile detection using supervised machine learning and in second phase we detect profile injection used for recommendations influence using outlier detection.
15 April 2022 in International journal of health sciences
Due to the obvious unstable increase in information, the web is saturated with data, which makes the data search a complicated task. Existing web-based recommendation systems include shortcomings such as a lack of capability as well as scalability when dealing with online data, and blockages created by traffic while utilising the website during peak hours. Web recommendation systems help consumers find the right content and make the information search process easier. Web usage mining is regarded as the primary source for web recommendation, and it is used in conjunction with association rule mining and the C4.5 algorithm to recommend online pages to the user. The Google search engine has been widely enhanced the likelihood on the system's suggested structure. A web log is created when a user enters a search query into a search engine. This query would be compared to the web logs by the proposed system. The associate rule mining technique helps in matching the user's search query to the online log. The C4.5 algorithm is linked to a priority based on reviews, which obviously ranks the search based on priority for greater validation result.