Rent Prediction Models with Floor Plan Images @IEEE GCCE 2019

A rent prediction model with floor plan image (FPI)
is proposed in order to evaluate that the image contributes to rent prediction accuracy. The proposed model calculates image feature vectors from FPIs using principal component analysis, combines them and attribute vectors, and applies a regressor. Mean squared errors (MSEs) between predicted and ground truth rents of the proposal are measured using 1,089,090 rental properties of the LIFULL HOME’S dataset. The experimental result suggests that the proposed model with linear regression and FPI slightly improves MSEs compared with the without FPI models; however, the differences are not significant and it.

Transcript

1. / 20 IEEE GCCE 2019 2019.10.16 Rent Prediction Models with Floor Plan Images 1 Ryosuke Hattori¹ɹKazushi Okamoto¹ɹAtushi Shibata² ¹ Graduate School of Informatics and EngineeringɼThe University of Electro Communications ² Graduate School of Industrial TechnologyɼAdvanced Institute of Industrial Technology
2. / 20 IEEE GCCE 2019 2019.10.16 Features of properties • almost properties with different attributes • property attributes • age
   Ttory
   Cuilding structure, and so on • effect on prices Rental case comparison method (Onokiɼ2016) • referring around the target property and the similar properties • linear regression model Introduction 2 K. Ohno: Keizoku Chinryou Kantei Hyouka Wo Saikou Suru, Jutaku-Shimpo Inc., 2016 ˆ yi = ↵1x1 + ↵2x2 + ↵3x3 + · · · + ↵nxn + <latexit sha1_base64="vQx08F43bZHjET/G6F0shHJHj2A=">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</latexit> <latexit sha1_base64="vQx08F43bZHjET/G6F0shHJHj2A=">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</latexit> <latexit sha1_base64="Mhn41z7d/Q2n10aqYFv7sL5P/oA=">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</latexit> rent distance story age location ˆ y <latexit sha1_base64="xhpVBugrKGqsl/IOdSNq+waH9Pk=">AAACbXichVHLSsNAFD2N7/qqiiAoUixVV2WqguJKdOPSttYWq5QkTjWYJiGZFmroD7gWXIiCgoj4GW78ARf9BHHhQsGNC2/SgGhR7zAzZ87cc+fMjGLpmiMYa4SktvaOzq7unnBvX//AYGRoeMsxK7bKs6qpm3ZekR2uawbPCk3oPG/ZXC4rOs8ph2vefq7KbUczjU1Rs/huWd43tJKmyoKobXfnQBZurV6PFiMxlmB+RFtBMgAxBLFhRm6wgz2YUFFBGRwGBGEdMhxqBSTBYBG3C5c4m5Dm73PUESZthbI4ZcjEHtK4T6tCwBq09mo6vlqlU3TqNimjiLNHdste2QO7Y0/s49darl/D81KjWWlquVUcPB7LvP+rKtMscPCl+tOzQAlLvleNvFs+491CbeqrR6evmeV03J1mV+yZ/F+yBrunGxjVN/U6xdNnCNMHJH8+dyvYmksk5xNzqYXYymrwFd0YxxRm6b0XsYJ1bCBL5xo4wTkuQi/SqDQhTTZTpVCgGcG3kGY+AbAsjgA=</latexit>
3. / 20 IEEE GCCE 2019 2019.10.16 Related Works 3 • R P. Dahal, R K. Grala, J S. Gordon, L A. Munn, D R. Petrolia, J R. Cummings: A hedonic pricing method to estimate the value of waterfronts in the gulf of Mexico, Urban forestry & urban greening, vol.41, pp.184-194, 2019. • P. Deschermeier, B. Seipelt: A hedonic rent index for student housing in Germany, Cologne institute for economic research, pp.1–12, 2016. • Y. Jun, H. Kim: Measuring the effect of greenbelt proximity on apartment rents in Seoul, Cities, vol. 62, 2017. category of explanatory variable building structure location /access around environment sales contract image Ram+ 2019 ✔ ✔ ✔ Philipp+ 2016 ✔ ✔ ✔ Jun+ 2017 ✔ ✔ ✔ ✔ proposal ✔ ✔ ✔ ✔
4. / 20 IEEE GCCE 2019 2019.10.16 Floor Plan Standard 4 kitchen and other rooms kitchen square [0 m2,4.5 m2) [4.5 m2, 8.0 m2) More than equal to 
 8.0 m2 not separated Room 
 (R) separated Kitchen
 (K) Dinning Kitchen 
 (DK) Living Dinning Kitchen (LDK) DK LDK K R
5. / 20 IEEE GCCE 2019 2019.10.16 Floor Plan Images Floor plan standard is the same, but floor layout is different In japan, there is a custom to look at floor plan images 
 when searching for a desired rental property (Kiyota+ɼ2017) 5 Ex. Different floor plan images in the same property and floor standard :
   ,JZPUB
   5
   :BNBTBLJ
   )
   4VXB
   $
   4IJNJ[V
   3FBM
   FTUBUF
   BOE
   "*
   +PVSOBM
   PG
   UIF
   +BQBOFTF
   4PDJFUZ
   GPS
   "SUJpDJBM
   *OUFMMJHFODF
   WPM
   OP
   QQ
   
6. / 20 IEEE GCCE 2019 2019.10.16 Purpose: Validate influence of floor plan images to rent prediction Method • build prediction models with/ without floor plan images • compare prediction accuracy Prediction Models • Linear Regression (LR) • XGBoost • Support Vector Regression (SVR) Feature Extractor • Principal Component Analysis (PCA) 6 Our Approach
7. / 20 IEEE GCCE 2019 2019.10.16 Prediction formulaɿ Loss function : Prediction Model 7 ln <latexit sha1_base64="cip2YhSnDwuYWh3CEG2B1EDrAqU=">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</latexit> L(y, f(x)) = X (ln yi ln ˆ yi)2 = X (ln yi ˆ yi )2 <latexit sha1_base64="SVmYIXbXkVU3MysOFsi1FpuiMIA=">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</latexit> ln ˆ y = f(x) = ↵Tx + <latexit sha1_base64="P9YN1Afnk0PPWLWhmqBsXdXQnxM=">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</latexit> u <latexit sha1_base64="x42OIFZO63bltL4H6gjza3ReZFg=">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</latexit> <latexit sha1_base64="x42OIFZO63bltL4H6gjza3ReZFg=">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</latexit> <latexit sha1_base64="x42OIFZO63bltL4H6gjza3ReZFg=">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</latexit> <latexit sha1_base64="x42OIFZO63bltL4H6gjza3ReZFg=">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</latexit> vi ln ˆ y = f(x) = ↵Tx + <latexit sha1_base64="P9YN1Afnk0PPWLWhmqBsXdXQnxM=">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</latexit> x = [u, ✓(v)] <latexit sha1_base64="tbvrKtiF/xOXDGMbwaD2COlchy8=">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</latexit> ಛ ௃ ྔ ଐੑ ϕΫτϧ ௞ྉ Image Features Property Variables Rent ln ˆ y = f(x) = <latexit sha1_base64="P9YN1Afnk0PPWLWhmqBsXdXQnxM=">AAACt3ichVHLShxBFD12Xjp5OMaN4KbJoBiEoUYDmsDAYDYu5+GoYJuhuq1xGqsfdtcMjk3/QH4gC1cRJAR/wH02/oCL2WSbhCwNZJOFd3oa1Ih6m65777n33DpVZfrSDhVjvSHtwcNHj58Mj2SePnv+YjQ79nI19NqBJeqWJ71g3eShkLYr6spWUqz7geCOKcWaufO+X1/riCC0PXdFdX2x6fBt127aFlcENbJVQ7qR0eIq6sZxsTljmJ7cCrsOOX3vdTG6mhtc+i0ef4iMwNFXYv1acS/WZ3XDFIo3sjmWZ4npN4NCGuSQWtnLfoGBLXiw0IYDAReKYgmOkL4NFMDgE7aJiLCAIjupC8TIELdNXYI6OKE7tG5TtpGiLuX9mWHCtmgXSX9ATB1T7Ix9ZefslB2zX+zfrbOiZEZfS5e8OeAKvzH6caL2916WQ16hdcm6U7NCE4uJVpu0+wnSP4U14Hf2P53X3lWnoml2yH6T/s+sx77RCdzOH+uoIqoHyNADFP6/7pvB6ly+MJ+fq7zJlZbSpxjGJF5hhu57ASUso4w67XuC7/iBn9pbraE1tdagVRtKOeO4ZtruBQt5qyw=</latexit> ✓(v) <latexit sha1_base64="yBk8zvv0E/O9omSla+R64kZVUTs=">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</latexit> <latexit sha1_base64="yBk8zvv0E/O9omSla+R64kZVUTs=">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</latexit> <latexit sha1_base64="yBk8zvv0E/O9omSla+R64kZVUTs=">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</latexit> <latexit sha1_base64="1tqRKu4qqxHI8AQAeehg2wQidig=">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</latexit> <latexit sha1_base64="2CcQ3v4PnGrVgMig/t6jEP2FmtA=">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</latexit> <latexit sha1_base64="2CcQ3v4PnGrVgMig/t6jEP2FmtA=">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</latexit> <latexit sha1_base64="tBF/c/fvuprQv5Z62Zn40d3w18Y=">AAAChXichVG7ThtBFD0sCRAngIEmEs0KC0QKrGuEANGAoKHkZUDClrW7jPGIfWl3bMms3FLwAxRUIEURIm34AZr8AAWfgChBSkPB3fVKKLEgd7Vzz5y5586ZGdO3ZaiI7rq07g8fe3r7PmU+f+kfGMwODW+HXj2wRNHybC/YNY1Q2NIVRSWVLXb9QBiOaYsd83AlXt9piCCUnrulmr4oO8aBK6vSMhRTlaxe8muyEpVUTSijpU9GJdOz98Omw0lvtL5lKtkc5SkJvRMUUpBDGmte9gdK2IcHC3U4EHChGNswEPK3hwIIPnNlRMwFjGSyLtBChrV1rhJcYTB7yOMBz/ZS1uV53DNM1BbvYvMfsFLHON3SJT3Sb7qie3p+s1eU9Ii9NDmbba3wK4MnXzf//FflcFaovare9axQxXziVbJ3P2HiU1htfePo9HFzYWM8mqALemD/53RHN3wCt/FkfV8XG2eIH6Dw73V3gu3pfIHxOuWWltOn6MMoxjDJ9z2HJaxiDUXe9xg/8QvXWq82pc1os+1SrSvVjOCv0BZfALomleY=</latexit> <latexit sha1_base64="yBk8zvv0E/O9omSla+R64kZVUTs=">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</latexit> <latexit sha1_base64="yBk8zvv0E/O9omSla+R64kZVUTs=">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</latexit> <latexit sha1_base64="yBk8zvv0E/O9omSla+R64kZVUTs=">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</latexit> <latexit sha1_base64="yBk8zvv0E/O9omSla+R64kZVUTs=">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</latexit> <latexit sha1_base64="yBk8zvv0E/O9omSla+R64kZVUTs=">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</latexit> <latexit sha1_base64="yBk8zvv0E/O9omSla+R64kZVUTs=">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</latexit>
8. / 20 IEEE GCCE 2019 2019.10.16 LR1 XGBoost2 • max depth: 6, 8, 10 SVR3 Used Regressors 8 ln <latexit sha1_base64="cip2YhSnDwuYWh3CEG2B1EDrAqU=">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</latexit> ln ˆ y = wTh(x) + c <latexit sha1_base64="9NdTbvfNygiSlMxs/yuuDlK+3Ew=">AAACoXichVFNLyxBFD0aD8N7BhuJTceEkJdIjUiIkAgbdmMYJNPepLvVmI7qj3TXDKPTf8AfsLAiEZH3M97GlsTCTxBLEhsLd3o6EcRzK1X31Ln33LpVZXjCCiRjdy1Ka1v7j47OrlR3z89fvem+/o3ArfomL5iucP0tQw+4sBxekJYUfMvzuW4bgm8ae0uN+GaN+4HlOuuy7vFtW991rLJl6pKoUnpOE06oVXQZ1qNInQ81wxU7Qd0mp+5Hf0LNt9X1SK2MvYscROPqb9UspTNsgsWmfgbZBGSQWM5NX0DDDlyYqMIGhwNJWEBHQKOILBg84rYREucTsuI4R4QUaauUxSlDJ3aP1l3aFRPWoX2jZhCrTTpF0PRJqWKE3bJL9siu2F92z16+rBXGNRq91MkbTS33Sr1Hg2vP36ps8hKVN9V/e5YoYybu1aLevZhp3MJs6muHx49rs/mRcJSdsQfq/5TdsX90A6f2ZJ6v8vwJUvQB2Y/P/RlsTE5kCa9OZRYWk6/oxBCGMUbvPY0FLCOHAp17hitc40bJKCtKTsk3U5WWRDOAd6YUXwHgWaGG</latexit> <latexit sha1_base64="9NdTbvfNygiSlMxs/yuuDlK+3Ew=">AAACoXichVFNLyxBFD0aD8N7BhuJTceEkJdIjUiIkAgbdmMYJNPepLvVmI7qj3TXDKPTf8AfsLAiEZH3M97GlsTCTxBLEhsLd3o6EcRzK1X31Ln33LpVZXjCCiRjdy1Ka1v7j47OrlR3z89fvem+/o3ArfomL5iucP0tQw+4sBxekJYUfMvzuW4bgm8ae0uN+GaN+4HlOuuy7vFtW991rLJl6pKoUnpOE06oVXQZ1qNInQ81wxU7Qd0mp+5Hf0LNt9X1SK2MvYscROPqb9UspTNsgsWmfgbZBGSQWM5NX0DDDlyYqMIGhwNJWEBHQKOILBg84rYREucTsuI4R4QUaauUxSlDJ3aP1l3aFRPWoX2jZhCrTTpF0PRJqWKE3bJL9siu2F92z16+rBXGNRq91MkbTS33Sr1Hg2vP36ps8hKVN9V/e5YoYybu1aLevZhp3MJs6muHx49rs/mRcJSdsQfq/5TdsX90A6f2ZJ6v8vwJUvQB2Y/P/RlsTE5kCa9OZRYWk6/oxBCGMUbvPY0FLCOHAp17hitc40bJKCtKTsk3U5WWRDOAd6YUXwHgWaGG</latexit> <latexit sha1_base64="9NdTbvfNygiSlMxs/yuuDlK+3Ew=">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</latexit> <latexit sha1_base64="9NdTbvfNygiSlMxs/yuuDlK+3Ew=">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</latexit> 1Linear regression class of scikit-learn (version: 0.20.2) 2https://github.com/dmlc/xgboost of python package (version: 0.82) 3SVR class of scikit-learn (version: 0.20.2) F = {f(x) = wq(x) } <latexit sha1_base64="vYZ5ogynMuxBxitBHt1d11Udu20=">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</latexit> ln ˆ y = K X k=1 fk(xi), fk 2 F, <latexit sha1_base64="kXKNmhiTJ1hFeMrJ2rKeQJCR2es=">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</latexit> ln ˆ y = ↵Tx + <latexit sha1_base64="n8nEqCpf1j5D7Rq0/4w8ipOkADU=">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</latexit>
9. / 20 IEEE GCCE 2019 2019.10.16 Experiment Environment LIFULL HOME ’S dataset (as of September 2015) • rental property data (70 variables, 5.33 million) • 120 x 120 pixel image data (83 million files) • used data • Tokyo • 1,089,090 properties • 34 property variables • floor plan image Computer (OS: Ubuntu 18.04.1) • CPUɿXeon(R) CPU E5-2650 v3 @ 2.30GHz • RAMɿ64GB 9
10. / 20 IEEE GCCE 2019 2019.10.16 10 Dataset (1/2) building structure/sales construct (19 variables) property type parking status num. of ground floors num. of underground floors floor num. property status available num. of parking num. of rooms building structure orientation of property purpose of use total num. of properties keeper exclusive area num. of vacant properties age of building immediate occupancy contract period days of occupancy the others: one-hot vector : continuous variables
11. / 20 IEEE GCCE 2019 2019.10.16 11 Dataset (2/2) location/acces (15 variables) city, ward, town city plan rail station 1 rail station 2 area description dist. to general hospital dist. to parking walking dist. to station 1 
 or bus stop walking dist. to station 2 
 or bus stop dist. to convenience store dist. to super market dist. to junior high school riding dist. to station 1 riding dist. to station 2 dist. to elementary school the others: one-hot vector : categorical each 80m grayscale Floor plan images
12. / 20 IEEE GCCE 2019 2019.10.16 Outline of Experiment 12 1. Dataset split • divide used dataset into 4 standards: K, R, DK, LDK • for each standard, divide 20% into the test dataset 
 and 80% into the development dataset 2. Hyper-parameter tuning • apply hold-out method to the development datasets • evaluation measure: Mean Squared Error (MSE) 3. Evaluation of LR, XGBoost, and SVR for the test datasets • MSE and learning time • visualize predicted values MSE = 1 N X (ln yi ˆ yi )2 <latexit sha1_base64="t0dz59RTfdsk2yjrs4h7PvtMieU=">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</latexit> <latexit sha1_base64="t0dz59RTfdsk2yjrs4h7PvtMieU=">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</latexit> <latexit sha1_base64="t0dz59RTfdsk2yjrs4h7PvtMieU=">AAAC23ichVHPS9xAGH1J1Vp/7dpeCr1IF0u9yEQKFaUglUIvirpdFXS7JOmowWwSktmF7bInb9Jr6aEnCyLin+Gl/0APQv+B0qOFXnrom9mAValOyMz3vW/e997MeEkYZEqIM8u+09Pbd7f/3sDg0PBIoTh6fzWLG6kvK34cxum652YyDCJZUYEK5XqSSrfuhXLN253X9bWmTLMgjt6oViKrdXc7CrYC31WE4uIsFlDGK4zhBf9NbCGFCx9tOOhwXjSzrmRooI4asw7zpzhEiIjZv5wW64FhbGKHmDJYJ0c7mMBbTHEdqBVLYlKYMXY9cPKghHwsxcUjtnyHmDLahqS0YhxSJOO3QbsCCbEqJV0aUpTUdanlyG1wl+QObXSX8zazjRyNmOuemWH7VAn5p+ag4+KbOBbn4qs4ET/En//2apse2kuLq9flyqRW2H9Y/n0rq24ua+eCdaNnxSufNl4Dek8Mok/hd/nN95/OyzMr4+0n4ov4Sf8H4kyc8gRR85d/uCxXPpsHcK5e9/VgdWrSYbz8rDT3Mn+KfjzCYz6/g+eYw2ssoULd75ZtDVpDdtXes/ftD92ttpVzHuDSsD/+BY2CoII=</latexit> <latexit sha1_base64="1tqRKu4qqxHI8AQAeehg2wQidig=">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</latexit> <latexit sha1_base64="W06s8Jmebi8e3jcSGo17fLbQvU0=">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</latexit> <latexit sha1_base64="W06s8Jmebi8e3jcSGo17fLbQvU0=">AAAC0HichVE9SxxBGH52zYfRSzytAmkkoiSNvGtjSBAEEWwS1PNU0Muxuxl1cW932Z07OI+r7CR9ilQJSAj5GWnyB1II/oGQ0kCaFHlmbsEkkmSGmXk/5nmfZ+YNsjgqtMiZ4w5du37j5vCtkdHK7Ttj1fHKZpG281DVwzRO8+3AL1QcJaquIx2r7SxXfiuI1VZwuGTyWx2VF1GabOhuphotfz+J9qLQ1wyl1Sd4ihqWMYkFrl3sIYePED146HN/ZneTKdBGC016ffoPcIoYCb1fMV3mI4vYxQFj2sb6ZbSPh3iOOZ4jzeqUzIodk1cNrzSmUI7VtPqOJV8gJY2RoUitacckKTh3KFeQMdYgpU9BmpQmrwwdsW3eUrxhhB5y36e3U0YT+qZmYdEhWWKu3D50Wj7Le7mQT/JBvsiPv9bq2RpGS5dnMMCqrDl2crf2/b+olv2sg0vUPzVrfvkjqzWi9sxGzCvCAb5z9Oqi9nh9ujcjb+Ur9b+RM/nIFySdb+Hpmlp/bRvg/fndV43NuVmP9ppgGPdwn133MI9FrGAVddKdO64z6lTchnvsngxa5Tplzybw23Bf/gSRb587</latexit> <latexit sha1_base64="hGIc4xcUXlFqjvduRZMaTIN6WK8=">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</latexit> <latexit sha1_base64="t0dz59RTfdsk2yjrs4h7PvtMieU=">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</latexit> <latexit sha1_base64="t0dz59RTfdsk2yjrs4h7PvtMieU=">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</latexit> <latexit sha1_base64="t0dz59RTfdsk2yjrs4h7PvtMieU=">AAAC23ichVHPS9xAGH1J1Vp/7dpeCr1IF0u9yEQKFaUglUIvirpdFXS7JOmowWwSktmF7bInb9Jr6aEnCyLin+Gl/0APQv+B0qOFXnrom9mAValOyMz3vW/e997MeEkYZEqIM8u+09Pbd7f/3sDg0PBIoTh6fzWLG6kvK34cxum652YyDCJZUYEK5XqSSrfuhXLN253X9bWmTLMgjt6oViKrdXc7CrYC31WE4uIsFlDGK4zhBf9NbCGFCx9tOOhwXjSzrmRooI4asw7zpzhEiIjZv5wW64FhbGKHmDJYJ0c7mMBbTHEdqBVLYlKYMXY9cPKghHwsxcUjtnyHmDLahqS0YhxSJOO3QbsCCbEqJV0aUpTUdanlyG1wl+QObXSX8zazjRyNmOuemWH7VAn5p+ag4+KbOBbn4qs4ET/En//2apse2kuLq9flyqRW2H9Y/n0rq24ua+eCdaNnxSufNl4Dek8Mok/hd/nN95/OyzMr4+0n4ov4Sf8H4kyc8gRR85d/uCxXPpsHcK5e9/VgdWrSYbz8rDT3Mn+KfjzCYz6/g+eYw2ssoULd75ZtDVpDdtXes/ftD92ttpVzHuDSsD/+BY2CoII=</latexit> <latexit sha1_base64="t0dz59RTfdsk2yjrs4h7PvtMieU=">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</latexit> <latexit sha1_base64="t0dz59RTfdsk2yjrs4h7PvtMieU=">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</latexit> <latexit sha1_base64="z5M9pnZi6P9Qvb0OzJ2kmuTf1lU=">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</latexit>
13. / 20 IEEE GCCE 2019 2019.10.16 Suitable Hyper-parameter for Each Model 13 floor plan standard hyper parameter K R DK LDK LR regularization method lasso lasso lasso lasso lambda 0 0.001 0 0.001 PCA 512 2048 1024 2048 XGBoost max depth 10 10 10 10 PCA 512 256 1024 1024 SVR epsilon 0.1 0.01 0.1 0.01 PCA 256 256 512 256
14. / 20 IEEE GCCE 2019 2019.10.16 Learning Time for the Training Dataset K 14 learning time [sec]
15. / 20 IEEE GCCE 2019 2019.10.16 MSE for the Test Datasets 15 # of data LR XGboost SVR w/ w/o w/ w/o w/ w/o K 30,446 0.028 0.029 0.015 0.016 0.159 0.161 R 94,420 0.024 0.029 0.013 0.014 0.099 0.100 DK 35,405 0.018 0.027 0.010 0.017 0.086 0.086 LDK 55,039 0.030 0.030 0.010 0.018 0.151 0.151
16. Predicted Values for the Test Dataset K XGBoost w/o XGBoost w/ LR w/o LR w/ SVR w/ SVR w/o
17. Predicted Values for the Test Dataset DK XGBoost w/o XGBoost w/ LR w/o LR w/ SVR w/ SVR w/o
18. / 20 IEEE GCCE 2019 2019.10.16 Cumulative Density (1/2) 18 K [30,446 properties] R [94,420 properties] = |1 − ̂ y y | error rate y : ground truth ̂ y : prediction ratio the number of data error rate
19. / 20 IEEE GCCE 2019 2019.10.16 DK [35,405 properties] LDK [55,039 properties] 19 Cumulative Density (2/2) = |1 − ̂ y y | error rate y : ground truth ̂ y : prediction
20. / 20 IEEE GCCE 2019 2019.10.16 Purpose: Validate influence of floor plan images to rent prediction Method • build prediction models with/ without floor plan images • compare prediction accuracy Result • Floor plan images reduce prediction variability • XGBoost with floor plan images archives the best MSE Future work: To apply Speeded-Up Robust Features (SURF) 20 Conclusion
21. 2-2. Visualize Predicted Values
22. / 20 IEEE GCCE 2019 2019.10.16 Analysis of Feature Importances on XGBoost 22 R K DK LDK Floor plan images Property variables
23. / 20 IEEE GCCE 2019 2019.10.16 MSE Evaluation BoF and PCA 23 # of data LR XGboost PCA BoF PCA BoF K 30,446 0.028 0.024 0.015 0.010 R 94,420 0.024 0.024 0.013 0.008 DK 35,405 0.018 0.019 0.010 0.010 LDK 55,039 0.030 0.031 0.010 0.010
24. / 20 IEEE GCCE 2019 2019.10.16 Cosine similarity of floor plan images (features is 10 dimension) 24 K DK LDK R The number of data The number of data The number of data The number of data
25. / 20 IEEE GCCE 2019 2019.10.16 RMSE evaluation and Learning time 25 # of data LR PCA-LR VGG-LR K 1,712 11,602 11,689 16,862 R 7,525 9,217 8,727 12,541 DK 3,548 11,156 10,957 16,467 LDK 6,842 20,121 19,646 22,878 # of data LR PCA-LR VGG-LR K 9,205 0.06 11 2,897 R 39,985 0.08 48 14,566 DK 18,975 0.05 23 6,471 LDK 36,595 0.07 45 13,026 RMSE for test data [yen] Learning time for train data [seconds]
26. / 20 IEEE GCCE 2019 2019.10.16 2-1ɽMSE evaluation and Learning time 26 # of data LR XGboost SVR w/ w/o w/ w/o w/ w/o K 30,446 0.028 0.029 0.015 0.016 0.159 0.161 R 94,420 0.024 0.029 0.013 0.014 0.099 0.100 DK 35,405 0.018 0.027 0.010 0.017 0.086 0.086 LDK 55,039 0.030 0.030 0.010 0.018 0.151 0.151 MSE for test data Learning time for train data [seconds] # of data LR XGboost SVR w/ w/o w/ w/o w/ w/o K 121,784 21.19 9 1.06 647.08 62.97 2,922.26 292.19 R 377,681 61.46 1.38 1,745.11 204.61 397,512.67 5,417.04 DK 141,622 23.79 1.12 658.53 66.32 6,480.69 597.36 LDK 220,154 35.98 1.47 1,018.97 111.00 13,120.27 1,297.53