Quantitive methods
Prediction of Prices for second-hand Car by Using Regression Models
Abstract:
Regarding the demand for cars all around the world, the need for the second-hand (used car) car market has been rising and creating a chance in business for both buyer and seller. The prices of the new cars in the industry are fixed and determined by the companies, but the cost of used cars is inconsistent. At the same time, many consumers want to sell their cars, they are not aware of the value of their property, and many of them cannot afford the price of the new car, so they prefer buying a second-hand vehicle.
The price of used cars is dependent on many features such as how old those vehicles are, miles traveled by car, etc. So, understanding which feature has the most effect on price is essential, and an expert system can be beneficial for predicting used car prices. Predicting the used car price is helpful for buyers and sellers because they know how much their property is worth. Used car sellers, online pricing services, and ordinary individuals are interested in this.
This data includes eight variables and 4809 records, and each record represents a car. First of all, in this project, we conducted a descriptive analysis that helps to find out the correlation between variables using the scatter plot of other features vs. price. After that, we try to find the best regression line with machine learning approaches to predict prices with the min absolute error (we use the train/test approach to evaluate the model). Linear regression and Random Forest are used in this project, and at the end of the project, we compare the results of the two methods and find the best predictor for price. The purpose of the project is: the purpose of the project is:
1-Which features have the most effect on the price.
2- Find the best predictor for price.
Diamond's price prediction: Linear Regression Model
Abstract:
Diamond is one of the rarest subsets of carbon with a crystal structure. The brilliance and sparkle of diamonds make them popular in the jewelry industry. Diamonds are scarce and expensive and have become popular because of their incredible ability to disperse light. [2]
Diamond is the hardest substance that is known today, so it is also used in industries and various machines for cutting, grinding wheels, and drilling, like giant tunneling machines.
As we discussed, diamonds are very useful in different industries. In 1953 system was developed for putting a value on diamonds based on their characteristics. These methods have many mistakes and errors because determining the price of diamonds has not been easy and depends on many variables like the stone's shapes, sizes, and purity. In this project, we want to find significant variables and compare their impact on price. After that, we test linear regression models to find the best and most efficient linear regression for predicting the price of diamonds.
The diamond dataset is downloaded from Kaggle, and it has about 54 thousand records and nine variables. After descriptive statistics and graphics analysis, we compare regression methods to find the best algorithm for evaluating the diamonds' price.
Prediction Of College Application
Abstract:
Universities need to be able to predict the number of applications so that they can plan their enrollment, allocate resources efficiently, and make well-informed decisions regarding admissions strategy. By projecting the number of applications, universities can allocate staff and facilities more efficiently and manage their budgets efficiently.
The collage number of application dataset, which has 18 variables and 777 observations, is used in this study. First, a descriptive analysis is carried out using diagrams like scatter plots, histograms, etc. Using approaches like linear regression, best subset selection, ridge regression, Lasso regression, random forest, bagging, Xg-boost regression, and gradient boosting regression, the prediction number of applications is calculated. In the end, various methods are compared and the most effective one is chosen.
Master's Thesis
Using Fuzzy Set Theory to Develop A New Estimation Method of Discrete Choice Models
Abstract:
Conjoint analysis was first developed in the marketing field to explore customer needs and preferences. Later, discrete choice experiments were used as an efficient alternative to conjoint theory. In discrete choice experiments, products are presented to the customer in a list of options, and the customer chooses the option that provides the greatest utility. After the survey results are collected, STSTS software is very useful for analyzing the data with multinomial logit models or conditional logit models. The results help us to find out which features of the product are most valuable to the customer.
In discrete choice experiments, the customer's ambiguity does not seem to be considered. For considering the ambiguity, fuzzy discrete choice models are developed in this dissertation. The fuzzy sets help to consider the uncertainty in the customer's choice. In this dissertation, discrete choice models are compared with fuzzy discrete choice models.
The financial case study is considered in this dissertation. Iranian stock market traders express their preference to buy company stocks by choosing options in a fuzzy way. Finally, the model is developed with the integration of discrete choice models and fuzzy sets to represent the uncertainty in customers' decisions.
Bachelor's Thesis
Prediction of loan requests by customers in financial institutions and banks
Abstract:
This project aims to predict the requests for loans by customers in financial institutions and banks. The paper is divided into three stages.
In the first stage, try to identify the missing data and find a suitable method to place them with less effect on the final answer.
In the second stage, the decision tree method was used to predict the request for loans by new customers in the future. Customers were classified into different branches, so these branches help to find out whether people with specific characteristics are suitable to obtain loans in the future or not.
In the third stage, the most effective variables for customer rating were identified through interviews with experts. Then, with the help of the AHP method, the criteria were weighted, and then, using the SAW method, a score from 1 to 10 was given to bank customers. This project uses various software such as Excel and Rapid Miner to understand the data and make predictions.
Work and time study
Work and time study in Amir Khan fast food
Abstract:
The purpose of this study was to investigate the process of making pizza, hot dogs, and hamburgers. In order to study the process, many charts were created that included the following components:
1) OPC
2) FPC
3) assembly chart.
Time was recorded three times with a stopwatch, and the average of these times was recorded as normal time. The study was conducted to evaluate the efficiency of the process and identify areas for improvement.