http://gottacouch.com
Project Guide: Dr. Hsinchun Chen Team Members: Steven Tagai Lin-Wei Tseng Jiayun(Melody) Zhao
Table of Content
Introduction ................................................................................................................... 3 Business Model .............................................................................................................. 4 Project Roles and Contribution ...................................................................................... 8 System Design and Architecture .................................................................................... 8 APIs and Data Sources.................................................................................................... 9 Web/Data Mining ......................................................................................................... 10 Novelty ......................................................................................................................... 14 Conclusion .................................................................................................................... 15
2
Introduction
At the end of each academic semester at The University of Arizona, many students move out of their living quarters due to graduation. Conversely, there are just as many, who move into their new living quarters as freshmen or those who move out to live with other colleagues and/or friends. Whatever the reason may be for moving, there is always a need for furniture and other essentials. And amongst the college demographic, there are certain necessities that most, if not all, college students must have in order to survive in a college setting. These necessities include desks, lamps, television sets, books, etc. Gotta Couch has been developed as a platform to provide students with a one-stop-search for these essentials to accommodate the need for essential furniture, electronics, and textbooks. With a special focus on college students, the system allows students to search for a selected and limited list of items that are essential to college life. This search feature pulls information from various sources that have items for sale that are both new and used, using Application Programming Interfaces (APIs) as well as web-crawling techniques to extract information. Amazon.com was used primarily to gather new items for sale. E-bay was used for both new and used items since both merchants and individuals sell items that are new and used. And finally, web-crawling methods were used to extract information from Craigslist for used items. Gotta Couch also provides a listing of Yard Sales that are upcoming, showing the exact locations of each respective sale on Google Maps. In addition, to make sure that no student ever goes yard sale shopping on a rainy day, Gotta Couch
3
provides forecast information provided by the National Oceanic and Atmospheric Administration. One key feature that Gotta Couch provides is the MeLinEven Television Bluebook. Similar to the Kelley Bluebook of automobiles, the MeLinEven Bluebook for televisions provides student shoppers with recommended prices for new and used televisions so that no student ever pays more than is necessary. This feature compiles a long list of television items, categorizes based on certain features and attributes, and provides recommended prices based on those attributes. While giving students an easy-to-use and easy-to-search feature for furniture, electronics, and textbooks, Gotta Couch can also potentially give sellers the space to advertise related items. With open areas for advertisements, Gotta Couch can generate revenue by displaying advertisements provided by Google Ad-sense or any other business that may want to pursue a partnership. By providing a portal for students to find what they need as well as recommend prices for televisions to prevent students from overpaying on an expensive item, Gotta Couch is a complete system that can be beneficial to both sellers and buyers while having the potential to generate revenue through substantial web traffic that it can possibly possess. The site can be accessed at: http://gottacouch.com/
Business Model
Target Customers:
The target customers for Gotta Couch will be college students that attend The University of Arizona. With students moving in and out of their living quarters every academic semester, many students
4
have the need to purchase furniture, electronics, and textbooks. Gotta Couch provides an easy-to-use and easy-to-search platform for students in need of essential items for surviving college life.
Services Provided:
Gotta Couch, as mentioned above, provides a portal for students to search for both new and used items that are necessary for college life. This narrowed list of items that students can search is based off of what students use the most in their daily lives. This list consists of the following: Furniture o o o Dining Room Furniture Living Room Furniture Bedroom Furniture
Electronics o o o o Televisions Lamps Microwaves Vacuums
Textbooks
The search results that are provided include new items from Amazon.com, new and used items from ebay.com, and mostly used items from craigstlist.org. The items from Craigslist that are displayed in
5
the search results are those items that are sold locally in Tucson since Craigslist provides listings based on geographical locations. Gotta Couch also provides yard sale locations to give students an even broader range of buying locations, maximizing the potential to save money through bargain shopping.
A key feature of Gotta Couch is the MeLinEven Bluebook of Televisions. Similar to the Kelley Bluebook of automobiles, the MeLinEven Bluebook recommends a reasonable price range for both new and used televisions. By collecting a sufficient amount of data regarding televisions and analyzing the prices based on the key attributes of television sets, the MeLinEven Bluebook can recommend students with the right price for the television they are interested in purchasing.
Competitors:
Regarding the functionalities, content and data sources, we mainly have two competitors who are Home and Garden and SPACIFY Furniture. Home and Garden focus on Home and garden furniture, furnishings, and accessories. It mainly uses Amazon EC2 and Shopping.com API. It includes a broad range of products from retailers around UK, DE, FR, AU and USA. SPACIFY Furniture is a designer dining room furniture site, which uses the Shopping.com API. Also, it includes modern and contemporary Italian dining room furniture and dining decoration. Compared with the above two websites, Gotta Couch has its own advantages:
Targeted Market: The primary users of Gotta Couch are university students while the targeted
6
customers of Home and Garden and SPACIFY Furniture are middle class families with higher income ranges. The purpose of the two competitors is not rested on providing living necessities but more on housing decoration. Both of them provide new stuffs with higher prices. Considering the financial situation of students and the fact that many of them only live in an apartment for several years, the combination of new and used stuffs with reasonable prices is a unique feature of Gotta Couch.
Combining various data sources: Gottacouch combines eBay, Amazon and Craigslist to provide more choices for students. Moreover, we integrated Google maps API for the Yard Sale functionality, which is an important novelty of our website. Besides furniture, which Home and Garden and SACIPY Furniture only provide, Gotta Couch users can also search for textbooks, yard sale locations and other household necessities, such as TV, microwave and lamp etc.
Price Recommendation: Home and Garden only provides a given price for each item, which can also be directly retrieved from eBay, Amazon, and shopping.com. Gotta Couch lists acceptable price ranges for televisions based on user preferred attributes.
Potential Business Partners and Financial Opportunities:
As a portal for students to find what they need, potential business partners would include any merchant or seller of items related to the items that can be searched by the Gotta Couch system. This includes any furniture, electronic, or book selling store. With the white space that surrounds the system’s interface, there is a huge potential for advertisement revenue. This space can either be sold
7
to the said merchants and sellers or can be used to display advertisements from Google Ad-Sense. With potential for high traffic, the use of multiple advertisement methods can be utilized for capital gains.
Project Roles and Contribution
Steven
Interface design/CMS Search function integration Amazon, eBay APIs
Lin-Wei
Web spidering and Parsing Google maps API
Melody
Database design Data mining
System Design and Architecture
Our project website is hosted on the Ecommerce Lab’s Apache server with a MySQL database to store the data we spidered from Craigslist and Amazon. Through our server, we make direct API calls to Amazon.com, eBay, and half.com to get useful information to display for our users. In order to retrieve the most relevant search results, our requests are usually fixed by a specific category ID and
8
a keyword. As for the data we retrieved from Craigslist, they are also spidered according to fixed search criteria and updated once every two weeks. Below is our website’s main architecture.
MySQL
Craigslist
Spider
Spidering and Data Mining
Apache gottacouch.com
User
Once a user visits Gotta Couch, they could start by searching for any furniture, electronics, textbooks, or yard sales. They could either search by the categories we provided or by typing in a specific keyword. For each search, our website returns the information directly from Amazon and eBay through API requests and, at the same time, display the information from Craigslist, which is fetched from our database.
APIs and Data Sources
APIs:
eBay APIs: With eBay API, we gathered the item name, current price, the remaining bidding time, and the link to eBay of new and used items that are in auction.
9
Amazon Associate Web Service: Through Amazon API, we gathered information such as the item name, ASIN, and the lowest price of “new” items that are currently sold on Amazon.com. For textbooks, we collected the book title, the author, ASIN, and its lowest price.
half.com by eBay: Our textbook information are also gathered from half.com API. Google Maps API: We used Google maps to visualize the location of all the yard sales in Tucson. Students who are searching for yard sales can provide a certain location and our web site will show all the yard sales that are located within a specific distance of that location.
NOAA (National Oceanic and Atmospheric Administration): The source of the weather forecast we provide to our users before they go yard sale shopping.
Spidering source:
Craigslist (tucson.craigslist.org): Craigslist provides information on different used items that are sold in Tucson and surrounding areas. The items that we spidered from Craigslist include furniture, televisions, lamp, microwaves, etc. For our furniture, we further categorized them into living room, dining room, and bedroom furniture.
Web/Data Mining
Web Mining:
One of Gotta Couch’s purposes is to provide students with not only new items, but also used items. Although eBay also has used items, we wanted to increase the amount of used items we could
10
provide and make it a special feature of our website. In order to collect enough data, we decided to spider tucson.craigslist.org. We used Offline Explorer to spider the web pages and wrote our own program to parse out the main information we needed. We spider Craigslist based on specific categories such as futon, dining room chairs, and lamps, etc. Our yard sale information is also spidered from Craigslist and the addresses were later on processed to work with Google maps. The data we collected are all stored in our database and updated once to twice every two weeks. For the data set we needed for our data mining, which is the MeLinEven TV Bluebook, we also collected information on new items from Amazon. We wrote our own program to spider and parse the content and stored the attributes we needed in our database.
Data Mining:
Gotta Couch is designed to provide living necessities to the University students. Price might be the most important determinant to students when purchasing a television because the price could vary in a big amount. Therefore, we want to give our users a recommendation of reasonable price ranges for different types of television.
Algorithm We use a decision tree, which is constructed with different attributes of a television, to generate the prediction. Decision trees represent a supervised approach to classification. A decision tree is a simple structure where non-terminal nodes represent tests on one or more attributes and terminal nodes reflect decision outcomes. The latest public domain implementation of decision
11
tree is C4.5. The Weka classifier package has its own version of C4.5 known as J48.
Data Collection and Edit Our data mainly comes from eBay, Amazon and Craigslist. Goods on eBay are used for biding, so the prices could not be used for training. We use the data from Amazon and Craigslist separately for new items and used items. To construct the decision tree, we need multiple attributes contributing to the final result. Therefore, our data parsing focused on TV sets which have more clear attributes such as types, screen size, brand and, color, and weight. The total number of data we used for training the model is about 3,500. Originally, we set color as an attribute, but we found out that it does not contribute to the variations of price when all the other attributes are the same. So we eliminated color from the decision tree. For the data of used TV from Craigslist, the attribute weight is unavailable in most cases, so we only used the other three attributes. The J48 algorithm requires all the attributes to be nominal value. Weka provides users with the filter function to discretize attributes with numeric value, so we divided screen size, weight, and price into different ranges. In order to make our recommendation more reasonable, we kept each interval of TV price within 20 dollars. For the brand, we found there were many unpopular brands, and each of them occupies very limited instances. In that case, only the brand could almost determine the price. So we picked the well known brands with the most instances and combined the rest together as “Other” brands.
12
Result For every TV set searched on our website, we provide a recommended price range to our users.
We compared the recommendation and the real price according to the same attributes repeatedly to guarantee the reliability. In addition, through our testing, we found something quite interesting. Regarding the HDTV, with the brand of Samsung and the screen size between 45 inches and 55 inches, the ones that weight between 62 lbs and 76 lbs actually have the highest prices. When the weight either goes up or goes down, the price will drop.
13
Novelty
Gotta Couch is currently targeted towards college students in Tucson to provide them with things that are necessary for their living and study. Our website integrated both the new stuffs and also used stuff and we provide the MeLinEven TV blue book to help them find televisions at the best price. The main novelties of our website are listed as follow:
Targeted Market: Our website is focused on users that are college students in Tucson Bringing together multiple sources: Users of our website can get all the item information from eBay, Amazon and Craigslist without searching them separately.
Search new and used items: All the items from Craigslist are used ones and the items from Amazon are new. eBay could contribute on both.
MeLinEven TV Blue Book: A reasonable price recommendation system to help our users find the right price for a particular type of television with different attributes.
14
Yard sale search with visualized map locations: Yard sale search is another very important novelty of our website which allows our users to find all the yard sales that are located within a certain distance based on where they live.
Conclusion
Gotta Couch provides a well integrated platform that combines information from various sources for college students to find living necessities. Students will be able to search for furniture, electronics, textbooks, and also nearby yard sale locations. To make Gotta Couch more attractive, we have a MeLinEven TV Bluebook which provides our users a best price for any specific type of television. In the future, we will expand the area in which we provide information from Tucson to the whole Arizona. In addition, we will also gain financial opportunities by partnering with item merchants and also by displaying advertisements form Google Ad-Sense.
15