The smart city is a recent concept that has gained popularity [1,2]. A lot of researchers are working on the development of smart cities because they offer various valuable services such as smart traffic, smart parking, smart telecom, etc. However, privacy, the intelligent use of large-scale big data, connectivity, efficiency, and infrastructure remain concerns in this area. To address these issues, emerging technologies and cyber-physical systems are employed. The cyber-physical systems (CPS) are a recent research area [3–5]. The CPS is an intelligent system and consists of sensors, controllers, and actuators. These smart devices work together and provide a connection between the cyber and physical world [6]. CPS integrates physical and computational components to monitor and control the physical processes seamlessly.

Fig. 1 shows the hierarchical example of CPS. In this figure, it is evident that cyberspace consists of clouds. Here, the cloud connects to network and storage servers. The servers are used for the computation and data storage. The physical sensing layer includes autonomous vehicles, power systems, public health, buildings, etc. These devices are connected to the cyberspace (data storage and computation services) [7]. The key objective of a smart city is to provide valuable services using cyber-physical space [8]. Recently, the global landscape has evolved into a hybrid environment. Thus the integration of cyber, physical, and social aspects is gaining popularity [9]. This phenomenon is known as Cyber-Physical-Social Systems (CPSS) [10,11]. The unique and most important aspect of CPSS from CPS is the existence of social domain. The purpose of a CPSS is to provide efficient, convenient, and personalized services topeople in smart cities [12,13].

Fig. 2 presents the basic structure and the key elements of a CPSS [14]. In this figure, a reader can see that cyber, physical, and social spaces are connected. The component of a social system is people. These people have relations with others and the relations are formed based on interaction, personal experiences, and observations. The second component is the ‘physical system’. This component comprised of sensors and actuators. Herein, the sensing devices are comprised of sensors actuators, etc. These devices are known as smart objects and are connected using communication technologies. The communication component is comprised of wireless and wired technologies. These are used to process data into the system and are shown in virtual space. The promising feature of a CPSS is that it provides an interface between different objects so that they can send and receive the data and as well carry out necessary actions. In general, most telecom companies record their user activity data including, SMS, calls, and internet for the monitoring purpose. This is commonly called Call Detail Record Data (CDR). Apart from calls, it contains various types of data; for instance; text messages and internet usage, etc.

The CDR is considered a gold mine for data scientists due to its usage and high potential [15]. When a person calls another using a mobile phone. In response, a CDR event was generated. In many cases, telecom operators store the data in the system’s database [15]. Due to the importance of CDR in customer analytics, there is a need to investigate this important aspect during planning in a smart city [16,17]. Since a typical telecom company has millions of subscribers and thus it generates a massive data [18,19]. Hence, how to handle the massive telecom data, storage, analysis, and processing remains a key concern. In addition to this, the telecom data should be clean, free from errors, with no duplication, and also fewer missing values [19]. Therefore, to address these issues, we propose a social cyber-physical system (CPSS) for analyzing and modeling large internet usage data. Our proposed system extracts useful information using CDR data and identifies healthy predictability of spatiotemporal patterns within the network traffic. Initially, our proposed model extracts useful information from raw CDR data and then identifies a healthy predictability of spatiotemporal patterns [20] within the network traffic. Briefly, our proposed model is comprised of three levels and each level has different functionality. Initially, the raw Call Detail Data (CDR) data is collected using a data collection layer. Here, preprocessing and error removal processes are performed. Then our proposed system passed it to the next level. In Data processing, CPSS performs pre-processing and data storage functions. Here, our proposed system constructs a graph and performs analysis on it. Briefly, the modeling processis performed, followed by the analysis, and thus the spatiotemporal analysis is conducted. Finally, our proposed CPSS system accurately identifies different areas of internet peak usage in the different regions of a smart city (we consider Milan).

In this section, we present a review of the most recent literature in this area. Sobb et al. [21] presented a review study. They first highlighted the growing and evolving state of CPSS. Then, reviewed CPSS especially the integration of emerging technologies named the Internet of Things (IoT), big data, machine learning, etc. The research gaps and the challenges were discussed in detail. Similar work has been carried out by Yin et al. [22]. In this study, they presented seven layered frameworks (7C), named D-CPSS using data-driven scenarios. Herein, the authors discussed the most recent development of IoT and the integration of social and large-scale networks.

Amin et al. [23] proposed CPSS for the analysis of hotspots using a smart city scenario. The proposed system uses graph-based metrics for the identification of hotspots in a city. However, the accuracy and the robustness were not measured. Brdar et al. [24] presented a knowledge retrieval model using telecom data [24]. Herein, They suggested traditional centrality measures named closeness and degree centrality [25]. Seufert et al. [26] proposed a Wi-Fi-based hotspot model for smart cities [26]. Herein, the Top-ten Wi-Fi hotspots were identified using a public dataset. They concluded that angles and the gamma distribution can be maintained using minimum distance. Nattapon et al. presented good research on CDRs using a telecom dataset [27]. In this study, they presented a method to clean the large data using “filters to filter”. Later, this method is used to remove anomalies. Kasem Ahmad et al. [28] presented a framework named the churn prediction SNA model. In this model, they combine both big data and machine learning [28]. Onnela et al. presented an analysis of telecom data using CDRs [29]. For this, they have selected a large dataset consisting of millions of CDR data. They have used CDR and considered them as weighted graphs [30]. According to these authors, it helps the readers to understand the structure of a network.

From the above literature review, we made the following observations. First, the research topic is new and thus still evolving. Secondly, limited research has been carried out in this area. Third, in existing studies, a generic model and the framework were presented. Thus, to fill this gap herein, we presented an interactive multilevel system for the analysis and modeling of internet activity data. The complete description details of the proposed multilevel system model are given below.

Fig. 3 shows our proposed multilevel system. In this figure, a reader can see a three-level system design model. In this section, we explain the functionality of each level.

In this section, we performed detailed experimental data analysis and also initiated a detailed discussion on the achieved experimental results. Herein, we use ‘Network X’ [31] as is a popular network package used for the analysis of networks. Network X core package provides a complete data structure and is comprised of various algorithms used for directed and undirected graphs.

In this section, we first explain the dataset. Secondly, we present a detailed data analysis on pre-processed data and then our proposed system performs spatial-temporal analyses. Herein, we have used the “Telecom Italia Big Data Challenge” [32]. This data was provided by the Italian Telecom.

Data between Trento and Milan cities were collected and a description of the dataset is listed in Table 1. In this table, the dataset type, issuer, and size of the dataset are given. The key elements of this dataset are:

id1: It shows the number of squares in Milan/Trento.

We have used processed CDR data and then 1-day activity has been performed. In this activity, we used 3 h of raw data and then aggregated it. As a result of aggregation, the data is further divided into 8 different slots, and the time duration for 24 h.

Herein, we proposed an interactive CPSS system on a big data platform using telecom data. The interactive CPSS system performs analysis and the modeling of internet usage data provided by Telecom Italia. Our proposed system first converts the large-scale telecom data into a graph and then applies the person correlation and the autocorrelation for peak usage of internet activity in the different regions of Milan. The proposed CPSS model is divided into different levels and each level has different functionality. At first, the data collection receives raw telecom data. The next step is to pass through the next level, named data processing. The data processing performs different functions, for instance, processing, storage and analysis, etc. The suggested internet activity record measures were employed. Finally, our proposed CPSS system accurately identifies different areas of peak internet usage in Milan city. In the future, we will consider the security aspects of our proposed model.