#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""Réseau de neurones tabulaire ROBUSTE : embeddings + ResNet-MLP,
ensemble multi-seed, Huber loss, AdamW + cosine, grad-clip, early stopping.
5-fold CV out-of-sample, comparable aux autres modèles (seed=42)."""
import json, re, warnings
import numpy as np, pandas as pd
import torch, torch.nn as nn
from sklearn.model_selection import KFold
from sklearn.metrics import r2_score
warnings.filterwarnings("ignore")
CUR_YEAR=2026
dev="mps" if torch.backends.mps.is_available() else "cpu"
print("device:",dev)
SEEDS=[42,1,7]            # ensemble par fold
D_MAIN=256; N_BLOCKS=3; D_HIDDEN=512

# ======== reconstruction du dataset (identique v4) ========
specs={}
for l in open("data/processed/specs.jsonl",encoding="utf-8"):
    r=json.loads(l)
    if r.get("specs"): specs[r["uid"]]=r["specs"]
COND={"a_renover":0,"original":1,"bon":2,"renove":3,"refait_neuf":4,"neuf":5,"inconnu":2}
BOOL=["turnkey","recently_renovated","garage","pool","waterfront","finished_basement","fireplace","central_ac","new_construction","income_property","renovated_kitchen","renovated_bathroom","large_lot","quiet_area"]
def gfeat(uid):
    s=specs.get(uid);o={"g_"+b:0 for b in BOOL};o.update({"g_cond":2,"g_hardwood":0,"g_quartz_granite":0,"g_view_water":0})
    if not s: return o
    for b in BOOL: o["g_"+b]=int(bool(s.get(b)))
    o["g_cond"]=COND.get(s.get("condition","inconnu"),2)
    fl=s.get("flooring") or [];o["g_hardwood"]=int("bois_franc" in fl)
    o["g_quartz_granite"]=int(s.get("countertop") in ("quartz","granit","marbre"));o["g_view_water"]=int(s.get("view")=="eau");return o
def parse_area(s):
    if not s or not isinstance(s,str): return np.nan
    t=s.lower();m=("m2" in t) or ("m²" in t);n=re.sub(r"[^0-9.,\s]","",t).strip().replace(" ","").replace(",","")
    try: v=float(n)
    except: return np.nan
    return (v*10.7639 if m else v) if v>0 else np.nan
def cc(c):
    if not c or not isinstance(c,str): return None
    c=c.split("(")[0].strip();return re.sub(r"\s+"," ",c).lower().strip() or None
def ti(x):
    try: return int(x)
    except: return np.nan
AMEN={"pool":["swimming pool","piscine","pool"],"waterfront":["waterfront","bord de l'eau","navigable","lakefront","riverfront","au bord de"],"garage":["garage"],"fireplace":["fireplace","foyer","wood stove","poêle"],"central_air":["central air","air conditionné","climatisation","a/c","heat-pump","thermopompe","air exchanger"],"finished_bsmt":["totally finished","finished basement","sous-sol fini","basement"],"near_water":["near water","waterfront","navigable"],"cul_de_sac":["cul-de-sac","cul de sac"],"golf":["golf"],"ski":["ski"],"near_park":["park","parc"],"near_school":["school","école","cegep","university","université","elementary","high school"],"near_highway":["highway","autoroute"],"transit":["public transit","public transportation","transport"],"no_rear_neighbor":["no backyard neighbors","sans voisin arrière"]}
def sd(s):
    s=s.lower()
    if "ft" in s:
        f=re.search(r"([\d.]+)\s*ft",s);i=re.search(r"([\d.]+)\s*in",s)
        return (float(f.group(1)) if f else 0)+(float(i.group(1))/12 if i else 0)
    m=re.search(r"([\d.,]+)\s*m",s)
    if m:
        try: return float(m.group(1).replace(",","."))*3.28084
        except: return None
    return None
def ra(d):
    if not d or "x" not in d: return None
    a,_,b=d.partition("x");la,lb=sd(a),sd(b);return la*lb if (la and lb and 1<la<200 and 1<lb<200) else None
RF={"r_office":["office"],"r_family":["family room"],"r_workshop":["workshop"],"r_winecellar":["wine cellar"],"r_sunroom":["sunroom","solarium"],"r_mezzanine":["mezzanine"],"r_library":["library"],"r_laundry":["laundry"]}
def pr(rooms):
    o={"room_area":np.nan,"n_rooms":0,"has_primary_bed":0,"has_basement":0,"n_levels_r":0}
    for k in RF: o[k]=0
    if not rooms: return o
    ar=0.0;lv=set()
    for r in rooms:
        nm=(r.get("name") or "").lower();l=(r.get("level") or "").lower();o["n_rooms"]+=1
        if l: lv.add(l)
        if "basement" in l or "sous-sol" in l: o["has_basement"]=1
        if "primary bedroom" in nm: o["has_primary_bed"]=1
        for k,kw in RF.items():
            if any(w in nm for w in kw): o[k]=1
        a=ra(r.get("dimensions"))
        if a: ar+=a
    o["room_area"]=ar if ar>0 else np.nan;o["n_levels_r"]=len(lv);return o
data=json.load(open("data/processed/maisons_avec_prix.json",encoding="utf-8"));rows=[]
for d in data:
    p=d.get("property") or {};ad=d.get("address") or {};ft=d.get("features") or []
    blob=(" ".join(ft)+" "+(d.get("description") or "")+" "+(d.get("title") or "")).lower()
    yb=ti(p.get("yearBuilt"));age=(CUR_YEAR-yb) if (not pd.isna(yb) and 1700<=yb<=CUR_YEAR) else np.nan
    pt=p.get("buildingType") or p.get("type");st=p.get("storeys") or p.get("levels")
    try: st=float(str(st).replace(",","."))
    except: st=np.nan
    uid=f"{d.get('source')}:{d.get('listingId')}"
    rec={"uid":uid,"price":d.get("priceNumber"),"source":d.get("source"),"sqft":parse_area(p.get("squareFeet")),"lot_sqft":parse_area(p.get("lotSize")),"beds":ti(p.get("bedrooms")),"baths":ti(p.get("bathrooms")),"parking":ti(p.get("parkingSpaces")),"storeys":st,"age":age,"ptype":(pt or "Unknown").strip(),"city":cc(ad.get("city"))}
    for n,kw in AMEN.items(): rec[n]=int(any(k in blob for k in kw))
    rec["n_images"]=len(d.get("images") or []);rec["virtual_tour"]=int(bool(d.get("virtualTourUrl")))
    rec.update(pr(d.get("rooms")));rec.update(gfeat(uid));rows.append(rec)
df=pd.DataFrame(rows)
df=df[df["price"].notna()];df=df[(df["price"]>=50_000)&(df["price"]<=15_000_000)]
df.loc[(df["sqft"]<300)|(df["sqft"]>20_000),"sqft"]=np.nan
df.loc[(df["lot_sqft"]<200)|(df["lot_sqft"]>5_000_000),"lot_sqft"]=np.nan
df.loc[df["beds"]>15,"beds"]=np.nan;df.loc[df["baths"]>12,"baths"]=np.nan
df["ppsf"]=df["price"]/df["sqft"];lo,hi=df["ppsf"].quantile([0.005,0.995])
df=df[~(df["ppsf"].notna()&((df["ppsf"]<lo)|(df["ppsf"]>hi)))]
ov=df[df["sqft"].notna()&df["room_area"].notna()&(df["room_area"]>200)];ratio=(ov["sqft"]/ov["room_area"]).median()
re_=df["room_area"]*ratio;ok=df["room_area"].notna()&(re_.between(300,20000))
df["sqft_final"]=df["sqft"];fill=df["sqft_final"].isna()&ok;df.loc[fill,"sqft_final"]=re_[fill]
df["has_sqft"]=df["sqft_final"].notna().astype(int);df["age2"]=df["age"]**2;df["n_images_c"]=df["n_images"].clip(upper=60)
vc=df["city"].value_counts();keep=set(vc[vc>=20].index);df["city_fe"]=df["city"].where(df["city"].isin(keep),"__other__").fillna("__other__")
tc=df["ptype"].value_counts();df["ptype_fe"]=df["ptype"].where(df["ptype"].isin(set(tc[tc>=30].index)),"Other")
df=df.reset_index(drop=True);print("N=",len(df))

amen=list(AMEN.keys());rv=["n_rooms","n_levels_r","has_primary_bed","has_basement","r_office","r_family","r_workshop","r_winecellar","r_sunroom","r_mezzanine","r_library","r_laundry"]
gv=["g_cond","g_hardwood","g_quartz_granite","g_view_water"]+["g_"+b for b in BOOL]
num=["sqft_final","has_sqft","lot_sqft","beds","baths","parking","storeys","age","age2","n_images_c","virtual_tour"]+amen+rv+gv
# indicateurs de manquant (robustesse) pour les colonnes à NaN
missnum=["sqft_final","lot_sqft","beds","baths","storeys","age","age2"]
for c in missnum: df["miss_"+c]=df[c].isna().astype(np.float32)
num=num+["miss_"+c for c in missnum]
y=np.log(df["price"].values).astype(np.float32)
cats=["city_fe","ptype_fe","source"];card={}
for c in cats:
    u=sorted(df[c].astype(str).unique());m={v:i for i,v in enumerate(u)};card[c]=len(u)
    df[c+"_ix"]=df[c].astype(str).map(m).astype(np.int64)
Xnum=df[num].astype(np.float32).values
Xcat=df[[c+"_ix" for c in cats]].values.astype(np.int64)

class ResBlock(nn.Module):
    def __init__(self,d,dh,p):
        super().__init__()
        self.bn=nn.BatchNorm1d(d);self.l1=nn.Linear(d,dh);self.act=nn.GELU()
        self.dp1=nn.Dropout(p);self.l2=nn.Linear(dh,d);self.dp2=nn.Dropout(p)
    def forward(self,x):
        z=self.bn(x);z=self.act(self.l1(z));z=self.dp1(z);z=self.l2(z);z=self.dp2(z);return x+z

class ResNetTab(nn.Module):
    def __init__(self,ncard,nnum):
        super().__init__()
        self.embs=nn.ModuleList();tot=0
        for c in ncard:
            d=max(2,min(64,round(1.6*c**0.56)));self.embs.append(nn.Embedding(c,d));tot+=d
        self.embdp=nn.Dropout(0.1)
        self.bn0=nn.BatchNorm1d(nnum)
        self.proj=nn.Linear(tot+nnum,D_MAIN)
        self.blocks=nn.ModuleList([ResBlock(D_MAIN,D_HIDDEN,0.2) for _ in range(N_BLOCKS)])
        self.head=nn.Sequential(nn.BatchNorm1d(D_MAIN),nn.GELU(),nn.Linear(D_MAIN,1))
    def forward(self,xc,xn):
        e=[emb(xc[:,i]) for i,emb in enumerate(self.embs)]
        x=torch.cat([self.embdp(torch.cat(e,1)),self.bn0(xn)],1)
        x=self.proj(x)
        for b in self.blocks: x=b(x)
        return self.head(x).squeeze(1)

def train_one(tr,te,seed):
    torch.manual_seed(seed);np.random.seed(seed)
    med=np.nanmedian(Xnum[tr],axis=0);med=np.where(np.isnan(med),0,med)
    Xn=Xnum.copy()
    for j in range(Xn.shape[1]):
        col=Xn[:,j];col[np.isnan(col)]=med[j];Xn[:,j]=col
    mu=Xn[tr].mean(0);sg=Xn[tr].std(0)+1e-6;Xn=(Xn-mu)/sg
    ym=y[tr].mean();ys=y[tr].std()+1e-6;yz=(y-ym)/ys
    ntr=len(tr);perm=np.random.permutation(ntr);vc_=int(ntr*0.9)
    tri=tr[perm[:vc_]];vai=tr[perm[vc_:]]
    def T(a,dt): return torch.tensor(a,device=dev,dtype=dt)
    model=ResNetTab([card[c] for c in cats],Xn.shape[1]).to(dev)
    opt=torch.optim.AdamW(model.parameters(),lr=2e-3,weight_decay=1e-4)
    EPOCHS=200
    sched=torch.optim.lr_scheduler.CosineAnnealingLR(opt,T_max=EPOCHS)
    lossf=nn.SmoothL1Loss(beta=1.0)   # Huber : robuste aux queues
    Xc_t=T(Xcat,torch.long);Xn_t=T(Xn,torch.float32);y_t=T(yz,torch.float32)
    best=1e9;bstate=None;pat=0
    for ep in range(EPOCHS):
        model.train();idx=np.random.permutation(tri)
        for i in range(0,len(idx),512):
            b=idx[i:i+512]
            opt.zero_grad();p=model(Xc_t[b],Xn_t[b]);l=lossf(p,y_t[b])
            l.backward();torch.nn.utils.clip_grad_norm_(model.parameters(),5.0);opt.step()
        sched.step()
        model.eval()
        with torch.no_grad():
            vp=model(Xc_t[vai],Xn_t[vai]);vl=((vp-y_t[vai])**2).mean().item()
        if vl<best-1e-4: best=vl;bstate={k:v.clone() for k,v in model.state_dict().items()};pat=0
        else:
            pat+=1
            if pat>=20: break
    model.load_state_dict(bstate);model.eval()
    with torch.no_grad():
        pz=model(Xc_t[te],Xn_t[te]).cpu().numpy()
    return pz*ys+ym

kf=KFold(n_splits=5,shuffle=True,random_state=42)
pred=np.zeros_like(y)
for f,(tr,te) in enumerate(kf.split(Xnum)):
    ps=np.zeros(len(te),dtype=np.float64)
    for sd_ in SEEDS:
        ps+=train_one(tr,te,sd_)
    pred[te]=ps/len(SEEDS)     # moyenne d'ensemble
    r2f=r2_score(y[te],pred[te]);print(f"  fold {f+1}: R²={r2f:.4f} (ensemble de {len(SEEDS)})")
r2=r2_score(y,pred);err=np.abs(np.exp(pred-y)-1)
print("\n=== NN ROBUSTE (ResNet-tab, ensemble multi-seed) — 5-fold CV OOS ===")
print(f"R² OOS = {r2:.4f} | erreur médiane = {np.median(err)*100:.1f}% | ±15% = {np.mean(err<=.15)*100:.0f}% | ±25% = {np.mean(err<=.25)*100:.0f}%")
print("\nRappel (mêmes folds): OLS 0.628 | RF 0.678 | NN simple 0.711 | HistGBM 0.743")
np.save("data/processed/nn_robust_oos.npy",pred)