2.3.2.2 Machine Learning Process and Challenges
The learning process based on given input where data are not sufficient does not produce reliable and robust results [7]. Machine learning algorithms get stuck in local minimum or maximum easily due to a large amount of data that leads to a problem called overfitting [7]. Furthermore, machine learning algorithms fail to learn positions and states (classes) when the number occurrence in the data is far less than others. For instance, within the smart healthcare domain, when capturing data, the probability of having all desired classes looks low, and the number of rare diseases would be insufficient for learning.
2.3.2.3 Deep Learning Process and Challenges
Feasible smart cities are established by technology‐driven foundations, and their initiatives are on different branches and domains, in which each of them may requires systems with high‐performance computing resources and technologies. Such systems have pros and cons such as saving energy and reducing the air pollution and reducing diagnose time, but writing something bad. One of the most popular technologies used to tackle such huge amounts of data is DL that is a special algorithm within machine learning and is efficiently used to obtain required knowledge from the input data, extracting the patterns that govern the whole data and also classify them. Several research studies are successfully applying DL on smart cities [24], such as urban modeling for smart cities, intelligent infrastructure for smart cities, and smart urban governance. Here, we aim to focus on a particular application of DL that is smart mobility and transportation.
AI definitely pushed all the science a step forward by making the systems and processes of scientific inquiry as smart as possible, for example, autonomous transportation systems [24] in smart mobility. Making a decision about whether the object seen is a human being or not is challenging [45]. Object detection is one of the challenging issues in smart mobility that surely boosts and facilitates automation in transportation systems. Consequently, this positively enhances and improves smart mobility in smart cities. Researchers in [45] explored analysis of decent object detection solutions like DL [46]. The scientists leveraged a well‐known object detection system, namely, YOLO (You Only Look Once), which was developed earlier by Redmon et al. [47] and assessed its performance on real‐time data.
2.3.2.4 Learning Process and Emerging New Type of Data Problems
In this section, we address possible challenges and solutions in the world of data analysis. The first and foremost problem that researchers tackle is lack of data for rare classes within the dataset that are used to make a model. The less number of samples we have in the dataset, the higher chance of ignoring that sample while we train and make the model. To handle this problem, meta‐learning has come to play an essential role to make a model only using few samples. It has three important promises: ZSL, one‐shot learning, and few‐shot learning [7]. ZSL is a certain type of meta‐learning when a training dataset does not have any samples for classes, and we still can predict them during a test process. For instance, a research work [44] was conducted to not use any annotation for processing vehicles tracklets. This study established a route understanding system based on zero‐shot theory for intelligent transportation, namely, Zero‐virus, which obtains high effectiveness with zero samples of annotation of vehicle tracklets. Further, another research work [40] established a new technique, namely MetaSense, which is the process of learning to sense rather than sensing to learn. This process takes advantage of the lack of samples of classes by learning from learning rather than learning from samples.
The process of non‐annotation helps to work with data that lacks classes because typical machine learning algorithms fail to detect non‐annotation classes. Furthermore, sensing to learn helps new algorithms predict information that is completely lacking from the dataset itself. These advancements lead us closer to (or are part of) ZSL, which is critical for the advancement of Smart Cities.
The second promise is one‐shot learning, in which each epoch in a training phase has only one sample per each class that is taken by a DL algorithm or a combination of neural networks [7].
The third promises but not the least one is few (k)‐shot learning in which each epoch in a training phase has only few (k) samples per each class that are taken by a DL algorithm or a combination of neural networks [7].
2.3.3 Decision‐Making Problems in Smart Cities
Decision‐making problems are becoming challenging issue in smart cities where not only the problem itself but also other relevant problems in other aspects of smart cities need to be analyzed. Additionally, decision makers must depict the consequence of the decision they are going to make. Thus, decision‐making systems are needed in smart cities in which the systems take care of all issues within the connected networks and only some limited information is taken that would be enough make an optimum decision. In this section, we highlight the challenging decision making problems and solutions.
2.3.3.1 Traffic Decision‐Making System
Traffic decision‐making system, as known as intelligent transportation system (ITS), aims to detect traffic flow within a smart city and offer optimal solutions using proper big data analytics [48]. Assessing and analyzing this big data plays a pivot role in decision making systems that makes the process time‐ and cost‐efficient [37]. Researchers addressed some case studies within traffic decision‐making systems, such as road traffic accidents analysis [49, 50], public transportation management and control, and road traffic flow prediction.