Lutfun Nahar Lota, Assistant Professor, CSE (original) (raw)
Uploads
Papers by Lutfun Nahar Lota, Assistant Professor, CSE
Cornell University - arXiv, Nov 6, 2022
User-specific future activity prediction in the healthcare domain based on previous activities ca... more User-specific future activity prediction in the healthcare domain based on previous activities can drastically improve the services provided by the nurses. It is challenging because, unlike other domains, activities in healthcare involve both nurses and patients, and they also vary from hour to hour. In this paper, we employ various data processing techniques to organize and modify the data structure and an LSTM-based multi-label classifier for a novel 2-stage training approach (user-agnostic pre-training and user-specific fine-tuning). Our experiment achieves a validation accuracy of 31.58%, precision 57.94%, recall 68.31%, and F1 score 60.38%. We concluded that proper data pre-processing and a 2-stage training process resulted in better performance. This experiment is a part of the "Fourth Nurse Care Activity Recognition Challenge" by our team "Not A Fan of Local Minima".
With the increasing popularity of the web and mobile devices, mobile applications (Apps) are now ... more With the increasing popularity of the web and mobile devices, mobile applications (Apps) are now used for a wide range of purposes. As mobile experiences are becoming more user-centered, system design is entirely dependent on who is using the app. For the success of the application, developers need to be aware of the users’ concerns and expectations of the application. Existing research has demonstrated that user evaluations often raise issues with an app’s usability, dependability, performance, or aesthetic appeal. These issues should be seen as Non-Functional Requirements (NFRs) in order to develop an application that meets users’ needs in all aspects. Since user reviews are often short, unstructured, and written in informal language, it might be hard to label them according to the standards of NFRs. In this context, our paper proposes a DCNR (Detection and Classification of Non-functional Requirements) approach that fine-tunes transformer language models (BERT, RoBERTa) to perfor...
Adjunct Proceedings of the 2020 ACM International Joint Conference on Pervasive and Ubiquitous Computing and Proceedings of the 2020 ACM International Symposium on Wearable Computers, 2020
Sensor-based human activity recognition has become one of the challenging and emerging research a... more Sensor-based human activity recognition has become one of the challenging and emerging research areas. Several machine learning algorithm with appropriate feature extraction has been used to solve human activity recognition task. However, recent research mainly focused on various deep learning algorithms, our focus of this study is measuring the performance of traditional machine learning algorithms with the incorporation of frequency-domain features. Because deep learning methods require a high computational cost. In this paper, we used Naive Bayes, K-Nearest Neighbour, SVM, Random Forest and Multilayer Perceptron with necessary feature extraction for our experimentation. We achieved best performance for K-Nearest Neighbour. Our experiment was a part of "The 2nd Nurse Care Activity Recognition Challenge Using Lab and Field Data" followed by the team MoonShot_BD. We concluded that with proper feature extraction, machine learning techniques may be useful to solve activity recognition with a low computational cost.
Adjunct Proceedings of the 2020 ACM International Joint Conference on Pervasive and Ubiquitous Computing and Proceedings of the 2020 ACM International Symposium on Wearable Computers, 2020
Human activity recognition on sensor data plays a vital role in health monitoring and elderly car... more Human activity recognition on sensor data plays a vital role in health monitoring and elderly care service monitoring. Although tremendous progress has been noticed to the use of sensor technology to collect activity recognition data, recognition still remains challenging due to the pervasive nature of the activities. In this paper, we present a Convolution Neural Network (CNN) model by our team DataDrivers_BD in "The 2nd Nurse Care Activity Recognition Challenge Using Lab and Field Data" which is quite challenging because of the similarity among the tasks. On the other hand, the dissimilarity among the users patterns of working for a particular task. Since CNN can retrieve informative features automatically, it has become one of the most prominent methods in activity recognition. Our extensive experiment on nurse care activity recognition challenge dataset also achieved significant accuracy of 91.59% outperforming the existing state of the art algorithms.
International Journal of Information Technology and Computer Science, 2017
Spam SMSes are unsolicited messages to users, which are disturbing and sometimes harmful. There a... more Spam SMSes are unsolicited messages to users, which are disturbing and sometimes harmful. There are a lot of survey papers available on email spam detection techniques. But, SMS spam detection is comparatively a new area and systematic literature review on this area is insufficient. In this paper, we perform a systematic literature review on SMS spam detection techniques. For that purpose, we consider the available published research works from 2006 to 2016. We choose 17 papers for our study and reviewed their used techniques, approaches and algorithms, their advantages and disadvantages, evaluation measures, discussion on datasets and finally result comparison of the studies. Although, the SMS spam detection techniques are more challenging than email spam detection techniques because of the regional contents, use of abbreviated words, unfortunately none of the existing research addresses these challenges. There is a huge scope of future research in this area and this survey can act as a reference point for the future direction of research.
Code comments are a vital software feature for program cognition &software maintainability. For a... more Code comments are a vital software feature for program cognition &software maintainability. For a long time, researchers have been tryingto find ways to ensure the consistency of code-comment. While doingthat, two of the raised problems have been dataset scarcity and languagedependency. To address both problems in this paper, we worked on adataset creation made using C# projects; there are no annotated datasetsyet on C#. 9,310 code-comment pairs of different C# projects wereextracted from a data pool. 4,922 code-comment pairs were annotatedafter removing NULL, constructor, and variable. Both method-commentand class-comment were considered in this study. We employed twoevaluation metrics for the dataset, one is Krippendorff’s Alpha whichshowed 95.67% similarity among the rating of 3 annotators for all thepairs & other is Bilingual Evaluation Understudy (BLEU) to validateour human-curated dataset. A modified model from a previous study isalso proposed, which obtained 96.2% using the p...
Adjunct Proceedings of the 2020 ACM International Joint Conference on Pervasive and Ubiquitous Computing and Proceedings of the 2020 ACM International Symposium on Wearable Computers, 2020
Nurse care activity recognition is a new challenging research field in human activity recognition... more Nurse care activity recognition is a new challenging research field in human activity recognition (HAR) because unlike other activity recognition, it has severe class imbalance problem and intra-class variability depending on both the subject and the receiver. In this paper, we applied the Random Forest-based resampling method to solve the class imbalance problem in the Heiseikai data, nurse care activity dataset. This method consists of resampling, feature selection based on Gini impurity, and model training and validation with Stratified KFold cross-validation. By implementing the Random Forest classifier, we achieved 65.9% average cross-validation accuracy in classifying 12 activities conducted by nurses in both lab and real-life settings. Our team, "Britter Baire" developed this algorithmic pipeline for "The 2nd Nurse Care Activity Recognition Challenge Using Lab and Field Data".
Cornell University - arXiv, Nov 6, 2022
User-specific future activity prediction in the healthcare domain based on previous activities ca... more User-specific future activity prediction in the healthcare domain based on previous activities can drastically improve the services provided by the nurses. It is challenging because, unlike other domains, activities in healthcare involve both nurses and patients, and they also vary from hour to hour. In this paper, we employ various data processing techniques to organize and modify the data structure and an LSTM-based multi-label classifier for a novel 2-stage training approach (user-agnostic pre-training and user-specific fine-tuning). Our experiment achieves a validation accuracy of 31.58%, precision 57.94%, recall 68.31%, and F1 score 60.38%. We concluded that proper data pre-processing and a 2-stage training process resulted in better performance. This experiment is a part of the "Fourth Nurse Care Activity Recognition Challenge" by our team "Not A Fan of Local Minima".
With the increasing popularity of the web and mobile devices, mobile applications (Apps) are now ... more With the increasing popularity of the web and mobile devices, mobile applications (Apps) are now used for a wide range of purposes. As mobile experiences are becoming more user-centered, system design is entirely dependent on who is using the app. For the success of the application, developers need to be aware of the users’ concerns and expectations of the application. Existing research has demonstrated that user evaluations often raise issues with an app’s usability, dependability, performance, or aesthetic appeal. These issues should be seen as Non-Functional Requirements (NFRs) in order to develop an application that meets users’ needs in all aspects. Since user reviews are often short, unstructured, and written in informal language, it might be hard to label them according to the standards of NFRs. In this context, our paper proposes a DCNR (Detection and Classification of Non-functional Requirements) approach that fine-tunes transformer language models (BERT, RoBERTa) to perfor...
Adjunct Proceedings of the 2020 ACM International Joint Conference on Pervasive and Ubiquitous Computing and Proceedings of the 2020 ACM International Symposium on Wearable Computers, 2020
Sensor-based human activity recognition has become one of the challenging and emerging research a... more Sensor-based human activity recognition has become one of the challenging and emerging research areas. Several machine learning algorithm with appropriate feature extraction has been used to solve human activity recognition task. However, recent research mainly focused on various deep learning algorithms, our focus of this study is measuring the performance of traditional machine learning algorithms with the incorporation of frequency-domain features. Because deep learning methods require a high computational cost. In this paper, we used Naive Bayes, K-Nearest Neighbour, SVM, Random Forest and Multilayer Perceptron with necessary feature extraction for our experimentation. We achieved best performance for K-Nearest Neighbour. Our experiment was a part of "The 2nd Nurse Care Activity Recognition Challenge Using Lab and Field Data" followed by the team MoonShot_BD. We concluded that with proper feature extraction, machine learning techniques may be useful to solve activity recognition with a low computational cost.
Adjunct Proceedings of the 2020 ACM International Joint Conference on Pervasive and Ubiquitous Computing and Proceedings of the 2020 ACM International Symposium on Wearable Computers, 2020
Human activity recognition on sensor data plays a vital role in health monitoring and elderly car... more Human activity recognition on sensor data plays a vital role in health monitoring and elderly care service monitoring. Although tremendous progress has been noticed to the use of sensor technology to collect activity recognition data, recognition still remains challenging due to the pervasive nature of the activities. In this paper, we present a Convolution Neural Network (CNN) model by our team DataDrivers_BD in "The 2nd Nurse Care Activity Recognition Challenge Using Lab and Field Data" which is quite challenging because of the similarity among the tasks. On the other hand, the dissimilarity among the users patterns of working for a particular task. Since CNN can retrieve informative features automatically, it has become one of the most prominent methods in activity recognition. Our extensive experiment on nurse care activity recognition challenge dataset also achieved significant accuracy of 91.59% outperforming the existing state of the art algorithms.
International Journal of Information Technology and Computer Science, 2017
Spam SMSes are unsolicited messages to users, which are disturbing and sometimes harmful. There a... more Spam SMSes are unsolicited messages to users, which are disturbing and sometimes harmful. There are a lot of survey papers available on email spam detection techniques. But, SMS spam detection is comparatively a new area and systematic literature review on this area is insufficient. In this paper, we perform a systematic literature review on SMS spam detection techniques. For that purpose, we consider the available published research works from 2006 to 2016. We choose 17 papers for our study and reviewed their used techniques, approaches and algorithms, their advantages and disadvantages, evaluation measures, discussion on datasets and finally result comparison of the studies. Although, the SMS spam detection techniques are more challenging than email spam detection techniques because of the regional contents, use of abbreviated words, unfortunately none of the existing research addresses these challenges. There is a huge scope of future research in this area and this survey can act as a reference point for the future direction of research.
Code comments are a vital software feature for program cognition &software maintainability. For a... more Code comments are a vital software feature for program cognition &software maintainability. For a long time, researchers have been tryingto find ways to ensure the consistency of code-comment. While doingthat, two of the raised problems have been dataset scarcity and languagedependency. To address both problems in this paper, we worked on adataset creation made using C# projects; there are no annotated datasetsyet on C#. 9,310 code-comment pairs of different C# projects wereextracted from a data pool. 4,922 code-comment pairs were annotatedafter removing NULL, constructor, and variable. Both method-commentand class-comment were considered in this study. We employed twoevaluation metrics for the dataset, one is Krippendorff’s Alpha whichshowed 95.67% similarity among the rating of 3 annotators for all thepairs & other is Bilingual Evaluation Understudy (BLEU) to validateour human-curated dataset. A modified model from a previous study isalso proposed, which obtained 96.2% using the p...
Adjunct Proceedings of the 2020 ACM International Joint Conference on Pervasive and Ubiquitous Computing and Proceedings of the 2020 ACM International Symposium on Wearable Computers, 2020
Nurse care activity recognition is a new challenging research field in human activity recognition... more Nurse care activity recognition is a new challenging research field in human activity recognition (HAR) because unlike other activity recognition, it has severe class imbalance problem and intra-class variability depending on both the subject and the receiver. In this paper, we applied the Random Forest-based resampling method to solve the class imbalance problem in the Heiseikai data, nurse care activity dataset. This method consists of resampling, feature selection based on Gini impurity, and model training and validation with Stratified KFold cross-validation. By implementing the Random Forest classifier, we achieved 65.9% average cross-validation accuracy in classifying 12 activities conducted by nurses in both lab and real-life settings. Our team, "Britter Baire" developed this algorithmic pipeline for "The 2nd Nurse Care Activity Recognition Challenge Using Lab and Field Data".