haojihu/tifuknn

kNN-based next-basket recommendation

tifuknn's Issues

Dataset

Hi!! I am facing issue for code execution for my dataset which has some new fields and that are not numbers so any suggestion from ur side how should I proceed!

Recommendation list

How to make a recommendation list for each customer?

nDCG calculation

nDCG的计算与我了解的不同。例如，top10推荐，最后取的推荐序列的前10个物品来计算DCG，取最后一个篮子中物品数来计算IDCG。（最后一个篮子物品数不固定）

def get_NDCG1(groundtruth, pred_rank_list, k): count = 0 dcg = 0 for pred in pred_rank_list: if count >= k: break if groundtruth[pred] == 1: dcg += (1) / math.log2(count + 1 + 1) count += 1 idcg = 0 num_real_item = np.sum(groundtruth) num_item = int(num_real_item) for i in range(num_item): idcg += (1) / math.log2(i + 1 + 1) ndcg = dcg / idcg return ndcg

temporal_decay_sum_history doesn't seems right

Hi @HaojiHu
Thank you for your amazing work! Really enjoy reading your paper! It was amazing!
I have just spotted some minor errors that may need your clarification.
In TIFUKNN.py line 595 (def evaluate)
it seems like temporal_decay_sum_history_training and temporal_decay_sum_history_chunk both have the same input.

Should the latter be changed to test_chink instead?

Thank you very much in advance!

Data processing scripts

Hi! Is it possible for you to provide the data processing scripts for the 4 datasets used?

First Costumer ID has empty Baskets set

When I use the preprocessing code, the generated Baskets for the first Costumer ID is empty [[-1]]; despite the original dataset, the first user has around 12 Baskets. Is it meant to be empty, or is there a problem?

model perform

how to improve recall value as paper describe?

Is the reported experiment on the paper using the same dataset & train/test split as on github?

Is the reported experiment on the paper using the same dataset & train/test split as on github?
Cause I ran command python TIFUKNN.py ./data/Instacart_history.csv ./data/Instacart_future.csv 900 0.9 0.7 0.9 3 10 but got inconsistent results from the paper reported.

[My output]
recall: 0.24572586831510687
NDCG: 0.3308917426465621

[Paper]
recall: 0.3952
NDCG: 0.3825