关于python:推荐系统基于物品的推荐算法实现

一. 实现思路

1. 筹备测试数据
2. 计算物品类似度, 应用 jaccard 计算类似度
3. 获取每个物品对应的类似物品
4. 获取最初的举荐数据

二. 代码实现

1. 筹备测试数据

users = ['User1', 'User2', 'User3', 'User4', 'User5',]
items = ['ItemA', 'ItemB', 'ItemC', 'ItemD', 'ItemE']
datasets = [[1, 0, 1, 1, 0],
    [1, 0, 0, 1, 1],
    [1, 0, 1, 0, 0],
    [0, 1, 0, 1, 1],
    [1, 1, 1, 0, 1],
]
df = pd.DataFrame(datasets, columns=items, index=users)

2. 计算物品类似度

items_similar = 1 - pairwise_distances(df.values.T, metric='jaccard')
items_similar = pd.DataFrame(items_similar, columns=items, index=items)

3. 获取每个物品对应的类似物品

topN_items = {}
for i in items_similar.index:
    _df = items_similar.loc[i].drop([i])
    _df_sorted = _df.sort_values(ascending=False)
    topN_items[i] = list(_df_sorted[:2].index)
print(topN_items)

运行后果

{'ItemA': ['ItemC', 'ItemE'],
 'ItemB': ['ItemE', 'ItemD'],
 'ItemC': ['ItemA', 'ItemB'],
 'ItemD': ['ItemE', 'ItemA'],
 'ItemE': ['ItemB', 'ItemD']}

4. 获取最初的举荐数据

rs_results = {}
for user in df.index:
    user_item = df.loc[user].replace(0, np.nan).dropna().index
    rs_result = set()
    for item in user_item:
        rs_result=rs_result.union(topN_items[item])
    rs_result -= set(df.loc[user].replace(0, np.nan).dropna().index)
    rs_results[user] = rs_result
print(rs_results)

举荐后果

{'User1': {'ItemB', 'ItemE'}, 
 'User2': {'ItemB', 'ItemC'}, 
 'User3': {'ItemB', 'ItemE'}, 
 'User4': {'ItemA'}, 
 'User5': {'ItemD'}}

残缺代码

import pandas as pd
from sklearn.metrics import jaccard_score
from sklearn.metrics.pairwise import pairwise_distances
import numpy as np
 
users = ['User1', 'User2', 'User3', 'User4', 'User5',]
items = ['ItemA', 'ItemB', 'ItemC', 'ItemD', 'ItemE']
datasets = [[1, 0, 1, 1, 0],
    [1, 0, 0, 1, 1],
    [1, 0, 1, 0, 0],
    [0, 1, 0, 1, 1],
    [1, 1, 1, 0, 1],
]
df = pd.DataFrame(datasets, columns=items, index=users)
# print(df)
items_similar = 1 - pairwise_distances(df.values.T, metric='jaccard')
items_similar = pd.DataFrame(items_similar, columns=items, index=items)
topN_items = {}
for i in items_similar.index:
    _df = items_similar.loc[i].drop([i])
    _df_sorted = _df.sort_values(ascending=False)
    topN_items[i] = list(_df_sorted[:2].index)
print(topN_items)
rs_results = {}
for user in df.index:
    user_item = df.loc[user].replace(0, np.nan).dropna().index
    rs_result = set()
    for item in user_item:
        rs_result=rs_result.union(topN_items[item])
    rs_result -= set(df.loc[user].replace(0, np.nan).dropna().index)
    rs_results[user] = rs_result
 
print(rs_results)
 
{'User1': {'ItemB', 'ItemE'},
 'User2': {'ItemB', 'ItemC'},
 'User3': {'ItemB', 'ItemE'},
 'User4': {'ItemA'},
 'User5': {'ItemD'}}