Recommend products based on historical queries of other users

Given the user data as in the following:

   user   query       date
0  jack   mango 2020-01-03
1  jack  banana 2020-01-04
2  jack   apple 2020-02-03
3  jack  orange 2020-03-03
4  john    meat 2020-07-03
5  john   water 2020-07-03

Now assume we have a new user enter mango, I am finding a good way to recommend user product.

One approach is the following based on item2vec:

import pandas as pd
df_user= pd.DataFrame( {'user':['jack','jack','jack','jack','john','john'],'query':['mango','banana', 'apple','orange','meat', 'water'],'date':['2020-1-3','2020-1-4','2020-2-3','2020-3-3','2020-7-3','2020-7-3']})
df_user['date']=pd.to_datetime(df_user['date'])

new_query='mango'

from gensim.models import Word2Vec

model = Word2Vec(sentences = df_user.groupby(['user'], as_index=False).agg(list)['query'], window = 9999999, min_count=1)
model.wv.most_similar(new_query, topn=10) 

Strangely, it gives

[('banana', 0.09904204308986664),
 ('orange', 0.004004828631877899),
 ('water', -0.022172965109348297),
 ('meat', -0.05908803641796112),
 ('apple', -0.1611100435256958)]

as output, where 'water' and 'meat' ranked above 'apple',

1. Is there any problem in my implementation?
2. Is there other good way to solve this problem instead of item2vec?