def get_deepfm(filepath, read_way="normal"):
if read_way == "normal":
xi, xv, label = ffmasvm2deepfm_v1(filepath=filepath, feat_len=444)
else:
deefmreade_out = DeefmReade(trainfile=filepath,
method="apply",
padding=True,
feat_len=444).ffm2deepfm()
xi, xv, label = deefmreade_out["feat_index"], deefmreade_out[
"feat_value"], deefmreade_out["label"]
dataset = BacthDataset(xi, xv, label)
return dataset
# dfm_params = {
# "use_fm": True,
# "use_deep": True,
# "embedding_size": 8,
# "dropout_fm": [1.0, 1.0],
# "deep_layers": [32, 32],
# "dropout_deep": [0.5, 0.5, 0.5],
# "deep_layers_activation": tf.nn.relu,
# "epoch": 30,
# "batch_size": 256,
# "learning_rate": 0.001,
# "optimizer_type": "adam",
# "batch_norm": 1,
# "batch_norm_decay": 0.995,
# "l2_reg": 0.01,
# "verbose": False,
# "eval_metric": roc_auc_score,
# "random_seed": 2017
# }
# 加载数据进来
filepath = "F:/kanshancup/def/FMdata/data/house_price/libffm.txt"
xi, xv, label = ffmasvm2deepfm_v1(filepath=filepath, feat_len=444)
print()
# 初始化模型对象
DFM_model = DeepFM() # 初始化对象
DFM_model.fit(xi, xv, label)
def get_deepfm(filepath):
xi, xv, label = ffmasvm2deepfm_v1(filepath=filepath, feat_len=444)
dataset = Bacth_dataset(xi, xv, label)
return dataset