A hybridization of deep learning techniques to predict and control traffic disturbances

Bechikh

et al. 2022

Neural Comput & Applic

Self Cite

Over the last decade, deep neural networks have shown great success in the fields of machine learning and computer vision. Currently, the CNN (convolutional neural network) is one of the most successful networks, having been applied in a wide variety of application domains, including pattern recognition, medical diagnosis and signal processing. Despite CNNs’ impressive performance, their architectural design remains a significant challenge for researchers and practitioners. The problem of selecting hyperparameters is extremely important for these networks. The reason for this is that the search space grows exponentially in size as the number of layers increases. In fact, all existing classical and evolutionary pruning methods take as input an already pre-trained or designed architecture. None of them take pruning into account during the design process. However, to evaluate the quality and possible compactness of any generated architecture, filter pruning should be applied before the communication with the data set to compute the classification error. For instance, a medium-quality architecture in terms of classification could become a very light and accurate architecture after pruning, and vice versa. Many cases are possible, and the number of possibilities is huge. This motivated us to frame the whole process as a bi-level optimization problem where: (1) architecture generation is done at the upper level (with minimum NB and NNB) while (2) its filter pruning optimization is done at the lower level. Motivated by evolutionary algorithms’ (EAs) success in bi-level optimization, we use the newly suggested co-evolutionary migration-based algorithm (CEMBA) as a search engine in this research to address our bi-level architectural optimization problem. The performance of our suggested technique, called Bi-CNN-D-C (Bi-level convolution neural network design and compression), is evaluated using the widely used benchmark data sets for image classification, called CIFAR-10, CIFAR-100 and ImageNet. Our proposed approach is validated by means of a set of comparative experiments with respect to relevant state-of-the-art architectures.

Section: Discussionmentioning

confidence: 99%

Joint design and compression of convolutional neural networks as a Bi-level optimization problem

Bechikh

et al. 2022

Neural Comput & Applic

Self Cite

“…Transportation studies can be considered as an intersection between several domains, methods, and techniques to propose, solve, and develop solutions for real-life problems. Artificial intelligence [1][2][3][4] and optimization [5][6][7] stand behind the majority of developed solutions. Among transportation and mobility problems, we recall the Vehicle Routing Problem (VRP).…”

Section: Introductionmentioning

confidence: 99%

Mixed Integer Linear Programming Models to Solve a Real-Life Vehicle Routing Problem with Pickup and Delivery

et al. 2021

Self Cite

This paper presents multiple readings to solve a vehicle routing problem with pickup and delivery (VRPPD) based on a real-life case study. Compared to theoretical problems, real-life ones are more difficult to address due to their richness and complexity. To handle multiple points of view in modeling our problem, we developed three different Mixed Integer Linear Programming (MILP) models, where each model covers particular constraints. The suggested models are designed for a mega poultry company in Tunisia, called CHAHIA. Our mission was to develop a prototype for CHAHIA that helps decision-makers find the best path for simultaneously delivering the company’s products and collecting the empty boxes. Based on data provided by CHAHIA, we conducted computational experiments, which have shown interesting and promising results.

“…Machine learning (ML) designs techniques have created models to forecast new data. 1 The key difference with the conventional method is that a system is designed from input data instead of a sequence of instructions. The unsupervised learning uses unlabeled-data, while supervised learning uses labeled-data.…”

Section: Introductionmentioning

confidence: 99%

“…2 ML was used to gather information and forecast future events in disciplines like industrial engineering, business, physics, computer engineering, medical, 3 pharmaceutical, and statistics, transportation. 1,[4][5][6] ML can play a significant role in predicting the price of a property with the recent growth in the property market. How to use algorithms to estimate land prices?…”

Section: Introductionmentioning

confidence: 99%

“…Therefore, designing a land price forecasting model will highly support forecasting future land values and defining property regulations. Machine learning (ML) designs techniques have created models to forecast new data 1 . The key difference with the conventional method is that a system is designed from input data instead of a sequence of instructions.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Price forecasting for real estate using machine learning: A case study on Riyadh city

Concurrency and Computation

Lahyani

Aldaej

et al. 2021

Self Cite

Real estate is potentially contributing to the economic growth. It has a strong correlation between property owners and beneficiaries. The accurate forecast of future property prices is particularly important. Therefore, real estate prices are reflecting the economic level of counties, and their price ranges are of great interest to both buyers and sellers. Developing a land-price forecast model could significantly assist in predicting future land prices and setting real estate regulations. In contrast, machine learning (ML) algorithms have demonstrated a great potential to perform predictions.Motivated by these assumptions, we develop in this article a set of ML algorithms to build models capable to increase the effectiveness of land price estimation. The ML algorithms adopted in this work include the decision tree, random forest (RF), and linear regression. We collected data from 5946 lands localized in the northern area of Riyadh, KSA. This data has been collected using GeoTech's DAAL website. The performance of the developed models has been assessed based on state-of-art performance metrics including mean absolute error, mean squared error, and median squared error.The experiments show that the RF based model outperforms the remaining models.