Publications

Decomposition Makes Better Rain Removal: An Enhanced Attention-Guided Image De-raining Using Deconvolutions Network

Downpour streaks noticeable all around show assorted qualities with various shapes, headings, densities, even the complex covered marvel, causing extraordinary difficulties for the de-pouring assignment. As of late, profound learning-based picture de-pouring techniques have been broadly examined because of their amazing exhibition. In any case, the majority of the current calculations actually have constraints in eliminating precipitation streaks while saving rich textural subtleties under muddled downpour conditions. To this end, we propose to disintegrate downpour streaks into numerous downpour layers and separately gauge every one of them along the organization stages to adapt to the expanding abstracts. To all the more likely portray downpour layers, a further developed non-neighborhood block is intended to abuse the self-similitude of downpour data by learning the all-encompassing spatial element relationships while lessening the estimation intricacy. Also, a blended consideration instrument is applied to direct the combination of downpour layers by zeroing in on the neighborhood and worldwide covers among these downpour layers. Broad tests on both manufactured blustery/downpour cloudiness/raindrop datasets, certifiable examples, the murkiness, and low-light situations show generous enhancements both on quantitative pointers and special visualizations over the present status-of-the-craftsmanship advances. | Article

An Analysis of Depression Detection Model Applying Data Mining Approaches Using Social Network Data

Depression, also defined as major depressive disorder, is a broad and straightforward psychiatric disorder that affects how we feel, experience, and respond. Fortunately, it is curable. Depression causes symptoms of depression and/or a loss of confidence in previously enjoyed interests. Any year, one out of every 15 people (6.7 percent) suffers from depression. Even though one out of every six people (16.6 %) will experience depression at any stage in their life. Depression can strike at any age, although it is most frequent between late adolescence and the mid-twenties. It is very difficult to locate individuals who suffer from depression. We revealed that social media delivers valuable signs for characterizing the appearance of depression in persons, as determined by a decline in social interaction, improved depressive effect, heavily clustered ego N/w, heightened relational and medicinal issues, and greater expression of religious participation. In this paper we analyze the depressing text; Manipulate data: Extract their features and categorize them using of principal component analysis, sentiment analysis approach, and build a predictor using cross-validate with Machine Learning models (Like Multinomial naïve Bayes, K nearest neighbors, and SVM).In which we have found a 99.7% Success rate with the use of a Multinomial naïve Bayes classifier. We suggest that our experiments and interventions can be useful in developing approaches for predicting the beginning of serious depression, either for healthcare agencies or on behalf of individuals, helping depressed people to be more diligent about their mental health.

Prediction of credit card defaults through data analysis and machine learning techniques

Bank of recent year plays a significant role in the development of the nation. The bank offers a few things that are directly dependent on any nation’s general economic and financial condition. Banking efficiency leads to the business, growth in the industry, economic growth, and support for the common man with savings, improving financial security. Bank Loan has been one of the fastest-growing financial services banks in recent years. However, with the increasing number of bank loan users, banks face an ever-increasing rate of bank loan decline. This program is offered primarily to a person or company of higher value than another. Under this scheme, a small amount can be provided as a cash transfer or electronic transfer to the debtor when they can be in demand. Few of them have not returned a set amount in time, so sometimes they do not. This situation creates a problem for the bank. Then with the help of historical data, the need to predict bank loan error can be determined. As such, machine learning may offer options for addressing the current issue and handling credit risk. This analysis has the function of forecasting the inability to pay the bank loan. The study found more than 10 million Bank of Taiwan records. Analysis of the logistic regression hits the relation between the class variable and the set of independent variables. The primary analysis produces exploratory views of data correctly. Further, this paper used ML algorithms to get predictions with accuracy to detect the default users based on transactional data. | Article