multiple regression

This article analyzes the demand for passenger air transportation in Iran. In analyzing passenger demand, the effect of geographic, socio-economic, and competitive factors has been explored. A data set on landings and takeoffs from the Kerman Airport has been collected for this purpose. An econometric model is first used to predict demand. The significance of all variables is investigated in this study. Once the variables that are of little significance are deleted, a new set of information is created. This information is then processed by the K-Means clustering algorithm and then used as training data for neural network learning. The neural network used for this analysis is an LSTM Deep Learning Network, which has been used to predict passenger demand for years. Finally, with economic and social variables including GDP, income, population, inflation, exchange rate, gasoline prices and oil prices for the coming years, the percentage change in the number of passengers for each year compared to the previous year has been predicted. Based on the inputs of the neural network, changes in air travel demand are determined based on the variables of gross national product, average Kerman income, gasoline price inflation, and oil price inflation. The most important variable is gross national product, which has a profound influence on air travel demand. The accuracy obtained in this method is 83%, which is a very good accuracy for air travel demand.

Various environmental conditions could have a significant effect on traffic flow. Understanding the effects of environmental conditions on a vehicle’s speed could provide an accurate estimation of travel speed and traffic management strategies adoption to prevent traffic crashes. In this paper, the speed of vehicles had been investigated based on different variables, such as the effect of some weather conditions (such as rainy, foggy, cloudy and sunny), air temperature, speed variations during nights and days, heavy vehicles percentages and vehicle volumes. In this research, the effects of these variables on vehicle speeds have been studied in order to decrease suburban crash frequencies, using a multiple linear regression model. Required data have been gathered from Iran Road Maintenance & Transportation Organization and Meteorological Organization Government agency in 2014. According to 7541 hourly samples at Tehran-Qom freeway, the results of the proposed model shows that rainy weather and night and day speed variation caused 1.355 and less than 1 kilometer per hour reduction in speed, respectively. Also, the proposed model has an acceptable and reliable error in speed prediction and the evaluation criteria such as mean absolute error, mean absolute percentage error and root mean square error were 3.579, 4.034 and 4.996, respectively.

The aim of this paper is to propose a model to forecast human and cargo air transport demand in Iran aviation industry. Using multiple regression method, the effects of geographical, economic, social, and competitive aspects were investigated within two separate models for passengers and cargo. Accordingly, we gathered data for 594 active air routes in Iran air network between 2009 and 2012 from Iranian airport holding company. One advantage of the proposed model in comparison to pervious researches is the exclusion of frequency of flight variable. This variable restricts the ability of forecasting the models due to its high correlation with demand for travel. Therefore, models that include with frequency of flight variable are not efficient for forecasting demand.
Additionally, the comparison of forecasting results and real demand for 2013 indicates the accuracy of forecasting model. Moreover, the results show that because of plenty of differences and erratic distribution between effective variables around the country, modeling should be specific to each city. This means that a general model for all cities is not appropriate and leads to an inaccurate forecasting, so we investigated separated models for metropolises of Iran to consider special situation of them.
Also, defining dummy (Binary) variables in modeling of special cities helps to increase fitting of model to real observations. We can incorporate special situation of some cities into model by means of these variables. Therefore, one of the methods to conquest the inconsistency between cities of Iran is applying dummy variables to simulate conditions of special cities.

Expansive soils are found in many parts of the world particularly in arid and semi-arid regions. They are considered a worldwide problem as they cause extensive damage to civil engineering structures. As a result، a clear understanding of the behavior of such soils is required for the effective design of structures and infrastructures on these soils. Researchers have demonstrated various methods to reduce expansive soils swelling potential، with one being soil treatment. In this paper an experimental program was developed to study the treatment of an expansive soil using mechanical and chemical techniques. Two different types of artificial fiber consisting of polyethylene (bar shaped) with a diameter of 0. 3mm and polypropylene (tape shaped) with a width of 3mm were used in order to randomly reinforce soil samples with various fiber contents of 0. 5%، 1% and 1. 5%; and fiber lengths of 10، 20 and 30mm. Chemical treatment was carried on using lime and cement as soil stabilizers with various percentages of 5%، 8% and 10%; for curing times of immediate، 3 and 7 days. Uniform and repeatable moist-compacted soil samples were prepared by static compaction at a chosen moisture content and dry density less than the optimum water content and maximum dry density from standard proctor compaction tests. Free swelling and swelling pressure tests were conducted using a one-dimensional oedometer apparatus. In addition multiple regression equations with 3 and 2 independent variables were developed based on experimental results. The proposed multiple regression equations present an understanding of the variation of free swelling and swelling pressure as a function of swelling time، fiber or chemical stabilizer content and fiber aspect ratio (fiber length divided to fibers diameter or width) or curing time for various values. Experimental results and statistical analysis indicated that soil treatment with artificial fibers and chemical stabilizers can indeed reduce swelling potential in expansive soils. Reduction of free swelling and swelling pressure in samples randomly reinforced with artificial fiber was seen to be dependent on fiber content، fiber length and fiber shape، where tape shaped fibers showed more reduction in free swelling and swelling pressure thus being more effective compare to bar shaped fibers. Furthermore reduction of free swelling in samples treated with chemical stabilizers were seen to be a function of percentage of chemical stabilizer and curing time with cement showing to be more effective compare to lime. In addition chemical stabilizers proved to be more effective in reducing soils swelling potential compare to artificial fibers. The suggested regression equations can predict and estimate the free swelling curve، final free swelling and swelling pressure for various swelling time، fiber or chemical stabilizer content and fiber aspect ratio or chemical stabilizer curing time with accuracy and a high degree of agreement between experimental and predicted values. Also، the regression equations can aid the performance of a sensitivity analysis which indicates the importance of each independent variable (fiber or chemical stabilizer content، fiber aspect ratio or chemical stabilizer curing time) in estimating final free swelling and swelling pressure.