Musabe Jean Bosco | Chongqing University of Posts and Telecommunications (original) (raw)

Musabe Jean Bosco

Musabe Jean Bosco was born in Huye, Rwanda, in 1985. He received a - Ph.D. and a master's in the research in Computational intelligence, Data mining, and knowledge discovery from Chongqing University of Posts and Telecommunications in China.
A dynamic, talented, international experienced, and committed manager. Extensive experience in Artificial Intelligent, Machine Learning, and Deep Learning with excellent knowledge in Data Science, data-driven, and Information knowledge discovery. My summary profile combines academic and professional experience in research interests such as Big data, data mining, granular computing, Multi-Granularity, cognitive computing, cloud computing, Fuzzy Algorithms, and machine learning algorithms. PhD Degree in Computer Science and Technology. I have more than 10 years of working experience as a research advisor and my research interests were based on the study of computational intelligence within the fields of Multi-granularity Remote Sensing, Land Change Science, Land Cover, and Land Use (LCLU), and Conservation. AI Environmental consultancy specifically in intelligent information processing, data analytics, designing, and managing rigorous, monitoring, training and education, data management, evaluation, and visualization system frameworks. Within such interdisciplinary teams, my particular strengths lie in the remote sensing of vegetation dynamics, land use and land cover change, and protected area management.
I am expertise covers the entire Computational Intelligent Framework (CIF) including exploratory research methods, sampling and measurement designs, research implementation, big data exploration, and visualization and quality assurance as well as qualitative analysis using python for big data analysis, deep learning, ArcGIS, ArcMap, and quantitative data analysis using the following statistical package software such as STATA and SPSS, questionnaire design and questionnaire programming in-field data collection software such as SurveyCTO, kobotoolbox, ODK, ONA, and CsPro, online field data collection monitoring and management, and visited – areas online mapping. I am expert in Data analysis using different methods including Apache Hadoop, Arche Spark, IoT, NoSQL, NewSQL Databases, Microsoft Azure HD Insight, HDFS, MapReduce, YARN, principal component analysis (PCA), multiple correspondence analysis (MCA), factor analysis (FA), cluster analysis, structural equation models (regression analysis (single and multivariate analysis), cross-tabulations and correlations, random forest (RF), support vector machine (SVM), decision tree (DT), multiple linear regression (MLR), logistic regression, k-nearest neighbor (K-NN), K-Mean (K-M), Naïve Bayes, hierarchical clustering, model selection and boosting such ad k-folder cross-validation, parameter tuning, gird search and XGBoost. I have been a consultant in many consultancies financed by Zhongzaisheng Environment Service Co., Chongqing Linkai Technology Co.
Phone: 0788355628
Address: Kigali
Rwanda
Africa

less

Uploads

Papers by Musabe Jean Bosco

International Journal of Advanced Engineering Research and Science, 2023

Abstract—Data-driven agriculture is a useful approach to producing good quality nutritious food s... more Abstract—Data-driven agriculture is a useful approach to producing good quality nutritious food sustainably without harming the planet. It aims to augment a farmer's knowledge with data and AI to eliminate guesswork and make informed decisions. The goal is not to replace farmers but to enhance their knowledge. Farmers possess valuable information about their farms, but many decisions are still based on guesswork. The ultimate objective is to remove the guesswork and replace it with data and AI. In this paper, the main objective is to design, manufacture and implement an adaptable smart system that Integrates Weather Data, Hardware, and Software for Real-time Data Integration in Irrigation and Soil Content Sensing for Optimizing Crop Production based on Soil Status. Secondly, our system must gather and analyse data to support informed decision-making, the last system must automate irrigation and fertilizer inputs to achieve precision agriculture. In summary, the paper emphasizes the importance of data-driven agriculture as a means to optimize crop production, enhance decision-making, and automate irrigation and fertilizer inputs for achieving precision agriculture.

https://ijaers.com/issue-detail/vol-11-issue-10/, 2023

A call center manages new and existing client questions and issues with the help of skilled spec... more A call center manages new and existing client questions and
issues with the help of skilled specialists. Existing customers answer
new consumers' inquiries and concerns. These questions may come
from new or existing clients. Call centers are important in Rwanda as
they enable companies to monitor calls. Companies can also analyze
their markets through data acquired through call centers. However,
setting up a call center is expensive. The running costs of a call center
are also large. Businesses that operate call centers spend a lot of
money running them, consequently reducing their profits. Therefore,
this study proposes a cheaper technique for handling call traffic
implemented using free PBX, which is a Linux based web-application
for monitoring call traffic. From the results of the simulations carried
out, a fast connection between mobile phones was observed. Moreover,
it was determined that the capacity of free PBX is unlimited, making it
ideal for use in call centers. The analysis shows that this project can be
implemented in different institutions on chipper prices. The existing
cost of implementing a call center on 50 users using hardware PBX is
100.000 USD, whereas, with the proposed solution using FreePBX
which is a Linux based web-application for monitoring call traffic, the
implementation cost can be between 5000 USD and 10000 USD with
the same range of users. The discounted price as compared to the
existing system can be estimated to be around 90%, which is much
cheaper.

Jurnal Kejuruteraan, Nov 29, 2023

There are numerous parking supervision and random booking procedures that regulate parking operat... more There are numerous parking supervision and random booking procedures that regulate parking operations. Travel time to the parking slot and walking time inside the terminus can still be dropped if the parker can book a precise parking spot instead of an arbitrary one. This is achieved by our proposal, called sPark which is an app-based parking method that includes indoor navigation facility i.e., an app for parking with indoor navigation facility. sPark's sharing system will rapidly book the optimal parking slots for parkers and advise them on the best feasible entrances for practice. Also, parkers will find the briefest path to their target using our proposed app's navigation technique, saving them a lot of time roaming to the building. Different parking methods like sPark (our proposed), non-directed and directed methods (existing) are designed and assessed. The designed and assessed simulation outcomes of sPark indicate an important decrease in the overall driving time by 30% to 60% as compared to the non-directed method which is an existing method. Additionally, the resource sharing module in our scheme i.e., an app for parking with indoor navigation facility called sPark has revealed a 9.99% decrease in driving time in comparison to directed methods (existing) that feature interior cruising and direction only.

Neural computing & applications, Feb 28, 2024

Deep convolutional neural networks (DCNNs) have emerged as powerful tools in diverse remote sensi... more Deep convolutional neural networks (DCNNs) have emerged as powerful tools in diverse remote sensing domains, but their optimization remains challenging due to their complex nature and the large number of parameters involved. Researchers have been exploring more sophisticated methodologies to improve image classification accuracy. In this paper, we introduce a multigranularity feature encoding ensemble network (MGFEEN) that is designed to fine-tune features at different levels of granularity. The network is trained in a two-step process: first, the output of granularity level i is used as the input for the next level; then, a fully connected layer is added to the pre-trained network to advance to the next level. The effectiveness of the MGFEEN's feature extraction is evaluated by feeding the globally extracted features to a softmax classifier for classification. By applying ensemble learning principles, our proposed MGFEEN achieves more accurate final predictions. We evaluate our model on three widely recognized benchmark datasets: UC-Merced, SIRIWHU, and EAC-Dataset. Notably, on the EAC-Dataset, our results show a significant 0.54% improvement in accuracy over a single-training-network setup, resulting in an impressive 98.70% accuracy level.

IEEE Access, 2021

Remote sensing scene classification is a fundamental responsibility of earth observation, aiming ... more Remote sensing scene classification is a fundamental responsibility of earth observation, aiming at identifying information granular for land cover classification. The multi-granular land use for multi-source remotely sensed image categories is now a principal task in remote sensing data augmentation and data selection. Understanding image representations are meaningful for the scene classification task. Training deep learning model-based approaches has to do with scene classification and brings about fantastic achievement. At the same time, these high-level approaches are computationally expensive and time-consuming. This paper introduces multi-granularity Neural Network Encoding architecture base on InceptionV3, InceptionReseNetV2, VGG16, and DenseNet201 architecture into remote sensing scene classification. To improve performance and to solve intra-class variation for multi-class scene issues remote sensing dataset. By using pre-trained CNN, activation function and ensemble learning have been adopting. InceptionV3 and VGG16 are used to extract features. InceptionResNetV2 use for fine-tuning, which consists of unfreezing the entre model and retraining the new data with a lower learning rate. The proposed fine-tune whole pre-trained model produces better results of test set up to 97.84 % than features extracted by InceptionResNetV2. Also, we use DCNNs ensemble average and weighted average to achieve better outcomes for 97.36% and 99.10%. In our experiments, we attempted to fine-tune the deep convolutional neural networks (DCNNs) training method for remote sensing scene classification on two public datasets UCM, SIRI-WHU, and one dataset I collected through the google earth engine from East Africa Community Countries (EACC) within nine classes within total 2112 labeled images. The results indicate that our proposed fine-tuning of the pre-trained model with few epochs and less computational time increases accuracy. INDEX TERMS Convolutional neural networks (CNNs), fine-tuning, granularity feature extraction, machine learning, and remote sensing (RS).

Deep learning classification has state-of-the-art machine learning approaches. Earlier work prove... more Deep learning classification has state-of-the-art machine learning approaches. Earlier work proves the deep convolutional neural network was successful and brilliant in different tasks such as image classification and image processing in remote sensing datasets. To recognize and clarify the physical aspect of the earth's surface land cover and exploit the land use is an exciting issue in environmental monitoring analyzing remote sensing data that is free and easy to get in different areas without time-consuming. To improve the quality of data sources and cooperating with image representation is still lack methods. We started with the traditional segmentation approach to divide an image into a single and compare multi-level into multiple segmentation for remote sensing imagery. We review the traditional machine learning and deep learning neural network classification tasks. We collected a dataset from six countries of Eastern Africa Communities with nine categories. We proposed an ensemble average model neural network with three models, combined as a single model trained, and achieved 96.5% validation accuracy. Finally, we compared state-of-the-art and extracted features pre-trained weight ImageNet and used traditional machine learning algorithms to improve accuracy.

NEURAL COMPUTING AND APPLICATION , 2023

Deep Convolutional Neural Networks (DCNNs) have emerged as potent tools in diverse remote sensing... more Deep Convolutional Neural Networks (DCNNs) have emerged as potent tools in diverse remote sensing domains, yet their optimization remains intricate due to their intricate nature and the extensive array of parameters involved. The enhancement of image classification accuracy has prompted researchers to delve into more sophisticated methodologies. In this particular investigation, we introduce a Multi-granularity Feature Encoding Ensemble Network (MGFEEN) designed to fine-tune different levels of granularity. The training of this network unfolds in a two-step process: commencing with the utilization of the output from granularity level i as the subsequent level's input, followed by the addition of a fully connected layer to the previously pretrained network to advance to the next level. The efficacy of this approach's feature extraction is assessed by subjecting the globally extracted features to a softmax classifier for classification. Through the application of ensemble learning principles, our proposed MGFEEN attains a more precise ultimate prediction. Our model is put to the test on three widely acknowledged benchmark datasets (UC-Merced, SIRIWHU, and EAC-Dataset). Notably, when employing the EAC-Dataset, our results exhibit a marked 0.54% enhancement in accuracy as compared to a singletraining-network setup, culminating in an impressive 98.70% accuracy level.

2021 IEEE International Conference on Computer Science, Electronic Information Engineering and Intelligent Control Technology (CEI), Sep 24, 2021

Software-Defined Networking (SDN) technology provides higher programmability and centralized mana... more Software-Defined Networking (SDN) technology provides higher programmability and centralized management in Industrial Internet of Things (IIoT)/industrial networks. Besides, it renders dynamic reconfiguration in order to optimize routing between controlled devices. Consideration of multiple metrics for Quality of Service (QoS) routing offers a precise network scheme and improves the performance compared to a single-metric. By applying the SDN platform, there is a possibility to build the forwarding paths based on multiple metrics such as delay, loss probability, and bandwidth, which are the paramount QoS requirements for many IIoT applications. This means that the performance is affected remarkably if the packet loss and delay exceed a definite limit and may become unsuitable for the endpoint. In this paper, an optimum path routing is used to ensure multiple constraints in Software-Defined-IIoT (SD-IIoT). With the use of this optimum path, the Reactive Flow Creation (RFC) approach increases the delay in the data transmission. This does not guarantee the delay-sensitive requirements for IIoT applications. To solve this issue, we propose dual solutions for flow creation: the optimized Proactive Flow Creation (PFC) approach and the combination of Reactive and Proactive Flow Creation (RPFC) approach. The improvement and capacity of the newly proposed approaches are demonstrated through the performances of reducing packet loss and delay with the consideration of both wired and wireless networks in a testbed setup.

Remote sensing is resource data accessible and easy to get in different areas without time-consum... more Remote sensing is resource data accessible and easy to get in different areas without time-consuming. The traditional image recognition task was unlimited to better classification. A convolutional neural network (CNN) was introduced to improve remote sensing image classification accuracy by eliminating the intra-class and class similarity. Training CNN from scratch requires a large annotated dataset that is occasional in the remote sensing area. Transfer learning of CNN weights from another large non-remote sensing dataset can occasionally help overcome typical RS image applications. Transfer learning consists of fine-tuning CNN layers to better the new dataset. In this paper, all of the experiments were done on nine categories for dataset collected in east Africa community countries (EAC) using three state-of-the-art architectures based on the effect of fine-tuning and pre-trained weights of CNN. Results indicate that fine-tuning the entire network is not always a significant way; we compared it with a process of using VGG16-DensNet pre-trained weights and RF as machine learning classified results can be improved up to 97.60. Alternatively, fine-tuning the top blocks can save computational power and produce a more robust classifier.

Deep learning classification is the state-of-the-art of machine learning approach. Earlier work p... more Deep learning classification is the state-of-the-art of machine learning approach. Earlier work proves that the deep convolutional neural network has successfully and brilliantly in different applications such as images or video data. Recognizing and clarifying the remote sensing aspect of the earth's surface and exploit land cover land use (LCLU). This article summarized the remote sensing emerging application and challenges for deep learning methods. We propose and examine four ways to learn efficient and effective CNNs to transfer image representation on the ImageNet dataset to recognize LCLU datasets. We use VGG16, Inception-ResNet-V2, Inception-V3, and DenseNet201 models pre-trained on the ImageNet dataset to extract features from the EAC dataset. We use pre-trained CNNs on ImageNet to extract features. The essential thing of our study is that we used principal component analysis (PCA) for feature selection to improve accuracy and speed up the model. We train our model by multi-layer perceptron (MLP) as a classifier. Lastly, we apply the multi-granularity encoding ensemble model. We achieve an overall accuracy of 92.3% for the nine-class classification problem. This work will help remote sensing scientists understand deep learning tools and apply them in large-scale remote sensing challenges.

2021 IEEE 2nd International Conference on Pattern Recognition and Machine Learning (PRML), 2021

Deep learning classification is the state-of-the-art of machine learning approach. Earlier work p... more Deep learning classification is the state-of-the-art of machine learning approach. Earlier work proves that the deep convolutional neural network has successfully and brilliantly in different applications such as images or video data. Recognizing and clarifying the remote sensing aspect of the earth's surface and exploit land cover and land use (LCLU). First, this article summarized the remote sensing emerging application and challenges for deep learning methods. Second, we propose four approaches to learn efficient and effective CNNs to transfer image representation on the ImageNet dataset to recognize LCLU datasets. We use VGG16, Inception-ResNet-V2, Inception-V3, and DenseNet201 models to extract features from the EACC dataset. We use pre-trained CNNs on ImageNet to extract features. For feature selection we proposed principal component analysis (PCA) to improve accuracy and speed up the model. We train our model by multi-layer perceptron (MLP) as a classifier. Lastly, we app...

2021 The 4th International Conference on Machine Learning and Machine Intelligence

Pattern Recognition and Machine Learning (PRML), IEEE International Conference, 2021

Remote sensing is resource data accessible and easy to get in different areas without time-consum... more Remote sensing is resource data accessible and easy to get in different areas without time-consuming. The traditional image recognition task was unlimited to better classification. A convolutional neural network (CNN) was introduced to improve remote sensing image classification accuracy by eliminating the intra-class and class similarity. Training CNN from scratch requires a large annotated dataset that is occasional in the remote sensing area. Transfer learning of CNN weights from another large non-remote sensing dataset can occasionally help overcome typical RS image applications. Transfer learning consists of finetuning CNN layers to better the new dataset. In this paper, all of the experiments were done on nine categories for dataset collected in east Africa community countries (EAC) using three state-of-theart architectures based on the effect of fine-tuning and pre-trained weights of CNN. Results indicate that fine-tuning the entire network is not always a significant way; we compared it with a process of using VGG16-DensNet pre-trained weights and RF as machine learning classified results can be improved up to 97.60. Alternatively, fine-tuning the top blocks can save computational power and produce a more robust classifier.

In 2021 The 4th International Conference on Machine Learning and Machine Intelligence (MLMI’21), September 17–19, 2021, Hangzhou, China. ACM, New York, NY, USA, 8 pages., 2021

2021 IEEE International Conference on Computer Science, Electronic Information Engineering and Intelligent Control Technology (CEI)

IEEE Access

Remote sensing scene classification is a fundamental responsibility of earth observation, aiming ... more Remote sensing scene classification is a fundamental responsibility of earth observation, aiming to identify information granular for land cover classification. The multi-granular land use for multisource remotely sensed image categories is now a principal task in remote sensing data augmentation and data selection. Understanding image representations are meaningful for the scene classification task. Training deep learning model-based approaches has to do with scene classification and brings about fantastic achievement. At the same time, these high-level approaches are computationally expensive and timeconsuming. This paper introduces multi-granularity Neural Network Encoding architecture base on InceptionV3, InceptionReseNetV2, VGG16, and DenseNet201 architecture into remote sensing scene classification. To improve performance and to solve intra-class variation for multi-class scene issues remote sensing dataset. By using pre-trained CNN, activation function and ensemble learning have been adopting. InceptionV3 and VGG16 are used to extract features. InceptionResNetV2 use for fine-tuning, which consists of unfreezing the entre model and retraining the new data with a lower learning rate. The proposed fine-tune whole pre-trained model produces better results of test set up to 97.84 % than features extracted by InceptionResNetV2. Also, we use DCNNs ensemble average and weighted average to achieve better outcomes for 97.3.6% and 99.10%. In our experiments, we attempted to fine-tune the deep convolutional neural networks (DCNNs) training method for remote sensing scene classification on two public datasets UCM, SIRI-WHU, and one dataset I collected through the google earth engine from East Africa Community Countries (EACC) within nine classes within total 2112 labeled images. The results indicate that our proposed fine-tuning of the pre-trained model with few epochs and less computational time increases accuracy. INDEX TERMS Convolutional neural networks (CNNs), fine-tuning, granularity feature extraction, machine learning, and remote sensing (RS) I.

Book Reviews by Musabe Jean Bosco

The new emerging remote sensing application brought deep convolutional neural