With the explosive growth in demand for mobile traffic, one of the promising solutions is to offload cellular traffic to small base stations for better system efficiency. Due to increasing system complexity, network operators are facing severe challenges and looking for machine learning-based solutions. In this work, we propose an energy-aware mobile traffic offloading scheme in the heterogeneous network jointly apply deep Q network (DQN) decision making and advanced traffic demand forecasting. The base station control model is trained and verified on an open dataset from a major telecom operator. The performance evaluation shows that DQN with traffic forecasting outperforms others at all levels of mobile traffic demands. Also, the advantage of accurate traffic prediction is more significant under higher traffic loads. INDEX TERMS Heterogeneous network, mobile traffic offloading, mobile traffic forecasting, deep learning, deep reinforcement learning, big data.

No abstract

Thermogravimetric analysis (TGA) testing was used to measure the change in weight of polished samples of Al-XSi (X = 0 and 1.2 mass%) alloys. The samples were heated at 843 K for 6 h in dry air or nitrogen gas. X-ray diffraction was used to monitor the formation of the oxide films on the surface of the samples. The surface oxide films were more compact after the Al alloy samples were heated in air, and the oxide films showed some cracks after being heated in nitrogen gas. The thermally formed surface oxide films on the Al-1.2 mass% Si alloy samples heated in air and in nitrogen gas possessed loose structures, which comprised mainly c-alumina, diaspore, and gibbsite, along with metallic silicon and/ or aluminum. The weight variation curve of the films appeared serrated; this can be attributed to chain reactions (3Si + 3O 2 ? 3SiO 2 + 4Al ? 3Si + 2Al 2 O 3 ) that occurred within the film.

Weight changes in Al-2 and 3.5 mass% Mg alloy samples were measured by thermogravimetric analysis. X-ray diffractometer tests were used to investigate the progressive development of thermally formed oxide films on the samples. Both samples were first heated in a dry air atmosphere. The oxide film formed on the Al-2 mass% sample comprised -alumina, MgO, MgAl 2 O 4 , and gibbsite. The film formed on the Al-3.5 mass% sample contained large quantities of MgO, but no MgAl 2 O 4 .The samples were then heated in a nitrogen gas atmosphere for <1:9 h. The Al-3.5 mass% sample contained greater amounts of gibbsite and -alumina than did the Al-2 mass% sample. The latter yielded a greater amount of AlN (and/or MgO). After an extended holding time ($6 h), the Al-3.5 sample contained a greater amount of AlN (and/or MgO), and its weight increased remarkably. The as-cast samples containing the Al-3.5 mass%Mg cube had a higher percentage of foggy film compared to those containing an Al-2 mass%Mg cube. Oxide films that are readily formed during melting tend to be trapped in aluminum alloy castings.

No abstract