We combine multiple machine learning algorithms, python crawling skills, Taiwan Open Datasets, Jeiba preprocessing, and 591 rent website to develop an Apartment Recommended System, which is called house for you. For the frontend visualization, we use HTML, JavaScript, CSS, and also provide statistical chart for data analysis. Through the web design and algorithm implement, the recommended system owns a high accuracy, which is evidenced by statistics called spearman correlation.
House for you provides the most suitable rented apartments according to the users' favor. Whenever you come to an alien environment, using "house for you" can solve what you are annoyed about living.
We take advantage of multiple datasets in the Taiwan open datasets at https://data.gov.tw/dataset/6564, and combine with the same address, which will lead into a complete data table and benefit of analysis. In this way, we can facilitate the analysis of the environment effective factor of the apartment, where the user is favor of but has no understanding yet. Quantify the factor and include into the cost–performance ratio we design for the prediction.
On the other hand, crawled the 591 (https://www.591.com.tw/) apartment-rented system using python until 3 May 2019, and 5362 records have been collected. For the preprocessing, the Jeiba to segment Chinese words and regular expression was implemented. After the work, we will get the internal effective factor of the apartment that renter providing, like the image in website, furnitures, and discount, etc. Likewise, Quantify the factor and include into the cost–performance ratio we design for the prediction.
First, take the two datasets mentioned above, use The World Geodetic System 1984 (WGS84) to convert the
Latitude and Longitude, and complete the combination of data that we use for analysis.
Second, consider the distance between the apartment and working place, we designed a center location method to find the most
suitable location for living.
Further, implement TFIDF and TFPDF in furnitures and Elo Rating System in the image aesthetic level for quantify into the cost–performance.
Final, combine the kmeans and scatter-gather algorithms to find the customized favor index for the cost–performance, and a complete apartment recommended
list will come into the view.
(For center location method)
-
Input working places or destinations (recommend a complete addredd)
-
Input time limits (you can accept) respectively of the places
(For customized favor index)
- Choose the apartment you prefer for 5 times.
After these, the recommended list will come up.
