5724128, 148 AM mT) in [2) out(2 mB) out} in a) outta Amport pandas as pd Smport matplotlib:pyplot as ple port Seaborn as sh port umgy 35. 9p sns.set_theme(color_codes-T-ue) af = pd.read_csv( Magicsricks.csv") Delhi House Price Predton -Jupyter Notebook ead) 4 soo 2 20 yg SRy Ret Sertor2heamwree 650 9 ga0q000 Randy r2movo NowPropany Bull Fear 6867.0 Data Preprocessing Part 1 Tech the numberof untque value fron ll of the object datatype loenity” a8 sor ect clam eco ts rater tte lt of stop) ahead “um 4 1ad 228 SenFimined 1.0 IDO et e-nove New Popa) fowtmat 8478 Exploratory Data Analysis locas 868nctbooksDehi House Pie Pr te5724123, 148 AM Delhi House Price Predton - Jupyter Notebook In [10]: |# List of categorical variables to plot Cat_vars = [Furnishing", "status", "Transaction’, “Type', “Bm, sathroon’, "Parking'] # create figure with subplots Fig, axs = plt-subplots or axs'= ans navel) | ncols=s, Figsize=(20, 28)) 1 create barplot for each cotegorical variable for 4, var in enunerate(eat_vars) sne.barplaticever, ye'Poice', data-df, ax-axs[i], estimator=op.nean) ans{ i). set_xticklabels(ars{] get_eticklabels(), rotation 4 adjust spacing between subplots Hie. tight_layoutO) 1 remove the etgth subplot Tigecelaxes(ox517)) fe renove the ninth subplot ‘f3gscelaees(axs()) 1 show pict show) | Tf por P| e u | L ae fd - all locas: 8888inotebooks/Delh House Price Preditonipynbt ae5724123, 148 AM Dei House Price Prediction -Jupyter Notebook in (12): |cat_vars = ['Furnishing’, ‘status", ‘Transaction’, "Type" 1t create a figure and oxes ¥5g, axe = plt-svbplots(orous-2, ncols-2, Figstze(20, 28)) # create a pie chart for each categorical variable for 4, var in enunerste(eat vars): AEC ten(axs. flat) # count the number of occurrences for each category cat_counts = dé{var].valuecourts() # create a pie chart Sx #lat(1].ple(cat_counts, Labele-cat_caunts.index, autopets "41.68%", startangle-92) set a title for cach subplot ax flat[1].set_title(#"(var} Distribution’) # adjust spacing between subplots 41g. tight layout) show the plat pit show) locahost 8888inotebooks/Delh House Price Prediction ipynbit ae5724123, 148 AM Delhi House Price Predton - Jupyter Notebook In (13): |mum_vars = ['Area', “Per sqrt] Fig, axs = plt.subplots(orows-1, neols-2, Figsize-(28, 10)) ax = axe. flatten’) for 4, var in enurerate(nun vars): shs.boxplot@cvar, datarde, 2x-aXs[1]) F3g.thght_Aayout() pit show) t i i i t i ' i i i i In (24): |mum_vars = ['Area', “Per sqrt") fig, axs = plt.subplots(orows=1, ncols=2, flgstze=(20, 18)) axe avs flatten’) for 4, var in enunerate nun vars): shs.violinplot(xovar, gata-df, ax-axs[i]) f34-thght_layout() plt shox) locas: 8888inotebooks/Delh House Price Preditonipynbt an5724123, 148 AM Delhi House Price Predton - Jupyter Notebook In (15): |mum_vars = ['Area', "Per sqrt] Fig, axs = plt.subplots(orows-1, neols-2, Figsize-(28, 10)) axe = axe. flatten’) for 4, var in enurerate(nun_vars): shs.scatterplat(acvar, yo! Price 1 huee'Furnishing’, datandf, F3g.thght_Aayout() pit show) 1 bes i In (18): |mum_vars = ['Area', “Per sare") Fig, axs = plt-subplots(orowse1, neols=2, Flgstze=(28, 10)) axe - avs flatten’) for 4, var An enunerate nun vars): shs.scatterplot(aevar, yo'Price’, huee'Status’, datasd®, ax-axs[i]) f34-thght_layout() plt shox) 1 a i locas: 8888inotebooks/Delh House Price Preditonipynbt site5724123, 148 AM Delhi House Price Predton - Jupyter Notebook In (17): |mumvars = ['Area', “Per sqrt] Fig, axs = plt.subplots(orows-1, neols-2, Figsize-(28, 10)) axe = axe. flatten’) for 4, var in enurerate(nun_vars): shs.scatterplot(acvar, yo'Price’, hue="Transaction’, datardf, ax-axs[i]) F3g.thght_Aayout() pit show) 1 al, 2 i in (28): /num_vars = ('Area', “Per_Sart"] Fig, axs = plt-subplots(orowse1, neols=2, Flgstze=(28, 10)) axe avs flatten’) for 4, var An enunerate nun vars): shs.scatterplot(xevar, o'Price’, hues'Type', datandf, axcaxs[]) f34-thght_layout() plt shox) 1 ole: : i locas: 8888inotebooks/Delh House Price Preditonipynbt ene5724128, 148 AM tn (19) out(19) 1n (20) out 29) tn (21) 1n (22) out 22) 1h (23) out 23} in (24) 1n (25) Delhi House Price Predton - Jupyter Notebook af. shope (2259, 19) check missing value check missing = df-tsnul1().sum() * 189 / dF. shape(] check missinglcheck missing > 0].sort_values(ascending-False) persqft —19.282176 parking 2.621128, Furnishing 0.397181, ype eso Sathroon —_@.158855 type: floates # FALL mull volue with nedton 4¢{ Por sats" ]-F111na(aPer_Saft’].median(), anplace=Teue) {FL ‘Parking’ J-f4llna(a{ Parking’ ].nedian(), inplacesTeue) 1 drop null value (» these 3 coluans because trhe null value {s below 2% {f.dropna(subsets[Furnishing’, "Type", ‘sathroon’}, inplacestrue) ‘af.nead) Ares BK Fora Statue Traneacton Type Per Soft @ mo00 8 «20 SarPune 00009 Reasy_pove New Prey ular hor 12515 4 1992-20 SemsFumined 1.0. 5080002 Rasey.te.move. Now Prpery —Aoanment S678 2 000 2 20. Fumred 1.0 15500000 Reatyterove Resale Aaurment 5657.0 3 oo 2 20 SemsFunined 1.0 «200002 Reasyto.rove Ree utter Fiber 50070 4 100 2 20 SamFumisned 1.0 5200000 Reaey.to.move Now Pronery BulKerFbo” 8657.0 af shape 1252, 18) Label Encoding for Object datatype # Loop over each column in the DatoFrane ahere dtype is “object for col in of. select_dtypes (includ printce{eol}: (@F[eolJ-unique())") the column nase and the unique vatues object" J)-coluans: Furnishing: ['Sent-Furntshed' ‘Fuendsheg’ “Unfurntshed' Status: ("Ready to_nove" "Alnost_rea ‘Transaction: ['New Property" Resale] ‘ype: [Bulider oor’ *apartnent’] 1 ‘from sklesrn inport preprocessing “# Loop over each column in the OstoFrane where dtype is ‘abject for col in 6f.select_dtypes(Include-[ ‘object’ J)-coluans: # initialize a LobetEncoder object Isbel encoder label encoder. #it(a#{ col] -unsque()) { rransform the column using the encoder f[col) = 1abel_encoser. teansfora(ef{¢ol]) reprocessing. LabelEncoder() it the encoder to the unique values in the column ¥ Point the colunn nave and the usique encoded values print(#{eol}: (aFeol|-unique)}") Furnishing: (1 @ 2] status: (1 6] Transaction: [@ 1] ype: [1 8] locas: 8888inotebooks/Delh House Price Preditonipynbt m65724128, 148 AM Delhi House Price Predton - Jupyter Notebook Correlation Heatmap In {26}: | rcorretotion Heotmap pit figure(Figsize-(22, 16)) Sns heateap(df.corr(), fnt=".2g", annot=true) caxessupplot:> Train Test Split in {28}: from sKlearn.nodel_selection inport train_test_split 4 perform tratn-test split Aavain, Ktest, yotrain, ytest © train test_split(efuérop("Pice’, arlsea), efLPrice'], test_sizer8.2, randon_stater Remove Outlier using IQR locahost 8888/notebooksIDoth House Price ane5724128, 148 AM Delhi House Price Predton - Jupyter Notebook tn [29]: |p concatenate X train and y_train for outlier renoval teain_af = pé.concat([X_train, y.traia], axis=2) # calculate the IQR values for each colum 1 = train ef quanttie(o.25) 03 + train af quantile, 75) raga # Renove outliers fran x trotn teain_af = train_of[-((erain df < (QL ~ 1.5 * 1QR)) | (train dé > (Q3 + 2.5 * 19R))).2ny(aetse2)) # Separate X train and y_troin after outlier renoval Xtrain = train_df.drop("Price", axisel) ylerain © enaimaafi Price’ Decision Tree Regressor In [30]: from skiearn.tree inport Oeeistontreetegressor fron sklearn-nodel_selection inport erdasearehcy From sklearn.datasets inport lead boston # create a Decistonireekegressor object dderee = Decistontreekegressor() “+ vefine the hyperparoneters to tune and their values paran_gria = ( saxdepth': (2, & 6 By minsanotes split’: {2, 4, 6. 8) ‘aincsanples_leef": [2,'2,'3,"4], naxfeatures': ['auto", “sat”, "1092", random. state": [@, 42] > # create a Grtasearchcy object arid_search = GridSearchcv{dtree, parangnid, c¥=5, scent vef_nean_squared_error') Fit the GrtdsearchC¥ object to the data eieseareh.FStQX train, y-train) 1 Print the best hyperparaveters print (arid_seareh.best-parans_) {nan depth’: 6, ‘max features’: “auto's “min_sanples_ + ‘min sanples_split": 2, ‘random_state': 42} In [31); from skiearn.tree Inport Oveistontreekegressor dtrge = Decisiontreehegressor(randon_stateni2, max depths, ax features dree. FRO train, y_teain) “auto', win_sanples_teaf=1, ain_sanples_split= ut[31}: Decssiontreetegressor(max depthe6, max Featuress'auto’, minsamples_leatea, Fandom. state-42) In [32]: from skearn saport metrics From sklearn.netrics Inport nean_absolute_percentage_erron Anport math yipred = dtree.predict(X test) ie = natrics.nean_absolute error(y_test, y_pred) rape = mean sbsolute percentage errar(y_test, pred) se = metrics.nean_squared_error(y_test, y_pred) Pa = netrics.r2_score(y test, ¥-pred) Paso = math. sgrtinse) print('MAE is ¢)' format(aae)) Prine mere fs ("format (nape)? print('MSE is ()' fornat(ase)) prine(’2 score is {)" -format(e2)) Prins(’RHSE score As ()"«Format(rase)) AE As 9490108, 759762005 ave is 0,42¢23957735847867 2 score is 0,4980663112331838 MSE score 42 22743687.609319054 locas: 8888inotebooks/Delh House Price Preditonipynbt one5724123, 148 AM Delhi House Price Predton - Jupyter Notebook In (33): imp ef = pd.bataFrane(( “Feature None": Xtrain.coluans, “tnportance"s dtree. feature saportances_ inp_d#.sort_values(by="Ingortance", ascending #42 = f4.nead(20) ple figure(Figsize=(20,8)) sns.barplet(dataefi2, x«' Importance’, ys'Feature Name") Plt -eitie('Feature Tnportance cach Accributes (Decision Tree Regressor)’, fontsizesi8) plt-xlabel ("inportance’, fontsize=16) plt.ylabel ("Feature Wane’, fontsize-16) pit show() Feature Importance Each Attributes (Decision Tree Regressor) = Pe mee Lai co | Bathroom Transaction Feature Name Furrishing Parking Status ao as o2 os os 05 06 or os Importance locas: 8888inotebooks/Delh House Price Preditonipynbt so5724128, 148 AM in (34): |amport shap explziner = shap.Treetxplainer(atree) Shap_values = explasnen-shap_valves(X_ test) Shap-sumary_plot(shap values, test) Delhi House Price Predton - Jupyter Notebook ih Area eee wore ote efene Per_Saft = + menalptnete Type we Bathroom . 3 Transaction + 2 Parking + 3 Furnishing “+ BHK + Status | ‘SHAP value (impact on model output) in [36]: explainer = shap.Explainer(dtree, x test) Shap_values = explaines(X-Lest) shap plots. waterfall (smap_values[@]) 999.51 Aix) = 10499 et_Saft 1300 = Area = Fumishing 1=Type 4=BHK 0 = Transaction 3 = Bathroom 1 = Parking 1 = Status 1.0 12 Random Forest Regressor locahost 8888inotebooks/Delh House Price Prediction ipynbit 167 Powe ) s125700.04 | arsesco +10915.98 10 14 16 nite5724128, 148 AM tn B7) 1m (38) out 38} tn 139) Delhi House Price Predton - Jupyter Notebook from sklearn.ensenble inport RandonforestRegressor from sklearn-nodel. selection dnport Gridsearchcy create @ Random Forest Regressor object P4- RandoarorestRegressor() 1 vefine the hyperparaneter grid paren grid = ( nex depth": (3s Sy Ty Se ‘aun_samples.spiit': f2, 5, 20), sincsamples leat": (1, 2,'8], fax features efsert Ty ‘tandon_state': [2, 42] > create a Grtdsearchev object arid_search © GridSearehCV{rr, paran_grid, cv=5, scoringe'r2") 1 FLe the GntdsearchOv object to the training data rid_search. Fit(X train, y_train) 1 Print the best hyperparaneters print("Best hypersaraneters: °, grid_search.best_parans_) est hyperparaneters: {°nax.cepth’: 9, ‘max features’: ‘sort', ‘min samples leaf": 1 spate a) from skleara-ensenble nport RandonForestRegressor P= RandenForestiegressor(randon state=42, max depth-3, min_sanples_splite2, min_sanples_leafs1, ax features "=a y PE ACOLRratn, y_train) andonForestRegressor(max depth=8, max featuress'sqrt", random state=42) from sklearn Snport metrics from skleara.netrics Inport nean_absolute_percentage_error Anport ath yipred = rf.predict(X test) ban - netrice nean_absolute_error(y_test, y_preé) ape = nean_absolute_percentage_errar(y. vest, y-pred) se = netrics.nean squared error(y test, y_pred) P= netrics.r2_score(y. test, ¥-pred) se = math. sqrtinse) print( MAE is {)' format (nae) print Mabe is ° format (mape)) print('MSE is ()' format(ase)) prine(’@2 score 5 {)--farmat(2)) prine(RHSE score ts ()".Foraat(rase)) aE is 9136272. 326071279 ave Ss 0,4378542709594508 2 score 1s, 0.4250026174:999723 NSE score {= 22413985,90799013 locahost 8888inotebooks/Delh House Price Prediction ipynbit ‘nin_sanples_split® 2, ‘random rae5724123, 148 AM Delhi House Price Predton - Jupyter Notebook In (40): imp ef = pd.bataFrane(( “Feature None": Xtrain.coluans, “tnportance" dtree. feature. iaportances_ » = Lap_df.sort_values(ty="Ingortance’, ascending #32 = f4.nead(20) ple #igure(Figsize=(20,8)) sns.barplet(dataefi2, x«' Importance’, ys"Feature Rane") Plt -titie('Feature Inportance cach Accributes (Random Forest Regressor)', fontsizesi8) plt-xlabel ("inportanee’, fontsize=16) plt.ylabel ("Feature Wane’, fontsize-16) pit show() Feature Importance Each Attributes (Random Forest Regressor) = Po mee Lai co | Bathroom Transaction Feature Name Furrishing Parking Status ao as o2 os os 05 06 or os Importance locas: 8888inotebooks/Delh House Price Preditonipynbt sa5724128, 148 AM Delhi House Price Predton - Jupyter Notebook in (41): |amport shap explziner = shap.Treetxplainer(rf) Shap_values = explasnen-shap_valves( test) Shap-sumary_plot(shap values, test) igh Atea Bathroom Per_Sqft BHK Parking Feature value Type Transaction oe Furnishing oo Status 1 4s oo os to ts ‘SHAP value (impact on model output) et in [42): explainer = shap.Explainer(ré, test, check saditivity-False) Shapivalues = explainer( test, check aaditivity=False) shapoplots.waterfall(smap_values[]) fix) = 1325121138 1 «Paring Usssts29 3 =Baneoom ED ‘= Type Beers 4 = Fumishing p zrse4 = BHK ~s604e 81 | 0 = Transaction | sse1759 1 = Status 0 locas: 8888inotebooks/Delh House Price Preditonipynbt sais