Winning solutions of kaggle competitions:

Generally at the end of every Kaggle competition, winners and other toppers share their solutions in the discussion forum. But it is a tedious task to search for the competitions and then for the solutions of these competitions when we need them. I always wanted to get the links to the solutions of all past Kaggle competitions at one place. I thought it would be a very good reference point many a times. Thanks to Kaggle team for the Meta Kaggle dataset, now I am able to get them in one single place through this notebook .

Thanks a lot to @sban for this wonderful kernel without which I wouldn't have got this notebook.

I have inlcuded only "Featured" & "Research" competitions and the competitions are ordered based on recency. The solutions are ordered based on the number of upvotes received.

Have fun and Happy Kaggling.!

P.S : There are few outlier discussions (which are not actually solutions) too ;)

P.S-2 : Thanks to my colleague @kazanova at h2o.ai for pointing me that this great solution has been missed out in the initial version. I I have added that now. In case if you find any of the solutions missing, please post them on the comments section with links. I will add them here.



In [ ]:

    
import numpy as np
import pandas as pd
from IPython.core.display import HTML

import warnings
warnings.filterwarnings("ignore")

data_path = "../input/"

competitions_df = pd.read_csv(data_path + "Competitions.csv")
comps_to_use = ["Featured", "Research", "Recruitment"]
competitions_df = competitions_df[competitions_df["HostSegmentTitle"].isin(comps_to_use)]
competitions_df["EnabledDate"] = pd.to_datetime(competitions_df["EnabledDate"], format="%m/%d/%Y %I:%M:%S %p")
competitions_df = competitions_df.sort_values(by="EnabledDate", ascending=False).reset_index(drop=True)
competitions_df.head()

forum_topics_df = pd.read_csv(data_path + "ForumTopics.csv")

comp_tags_df = pd.read_csv(data_path + "CompetitionTags.csv")
tags_df = pd.read_csv(data_path + "Tags.csv", usecols=["Id", "Name"])

def get_comp_tags(comp_id):
    temp_df = comp_tags_df[comp_tags_df["CompetitionId"]==comp_id]
    temp_df = pd.merge(temp_df, tags_df, left_on="TagId", right_on="Id")
    tags_str = "Tags : "
    for ind, row in temp_df.iterrows():
        tags_str += row["Name"] + ", "
    return tags_str.strip(", ")

def check_solution(topic):
    is_solution = False
    to_exclude = ["?", "submit", "why", "what", "resolution", "benchmark"]
    if "solution" in topic.lower():
        is_solution = True
        for exc in to_exclude:
            if exc in topic.lower():
                is_solution = False
    to_include = ["2nd place code", '"dance with ensemble" sharing']
    for inc in to_include:
        if inc in topic.lower():
            is_solution = True
    return is_solution

def get_discussion_results(forum_id, n):
    results_df = forum_topics_df[forum_topics_df["ForumId"]==forum_id]
    results_df["is_solution"] = results_df["Title"].apply(lambda x: check_solution(str(x)))
    results_df = results_df[results_df["is_solution"] == 1]
    results_df = results_df.sort_values(by=["Score","TotalMessages"], ascending=False).head(n).reset_index(drop=True)
    return results_df[["Title", "Id", "Score", "TotalMessages", "TotalReplies"]]

def render_html_for_comp(forum_id, comp_id, comp_name, comp_slug, comp_subtitle, n):
    results_df = get_discussion_results(forum_id, n)
    
    if len(results_df) < 1:
        return
    
    comp_tags = get_comp_tags(comp_id)
    
    comp_url = "https://www.kaggle.com/c/"+str(comp_slug)
    hs = """<style>
                .rendered_html tr {font-size: 12px; text-align: left;}
                th {
                text-align: left;
                }
            </style>
            <h3><font color="#1768ea"><a href="""+comp_url+""">"""+comp_name+"""</font></h3>
            <p>"""+comp_subtitle+"""</p>
            """
    
    if comp_tags != "Tags :":
        hs +=   """
            <p>"""+comp_tags+"""</p>
            """
    
    hs +=   """
            <table>
            <tr>
                <th><b>S.No</b></th>
                <th><b>Discussion Title</b></th>
                <th><b>Number of upvotes</b></th>
                <th><b>Total Replies</b></th>
            </tr>"""
    
    for i, row in results_df.iterrows():
        url = "https://www.kaggle.com/c/"+str(comp_slug)+"/discussion/"+str(row["Id"])
        hs += """<tr>
                    <td>"""+str(i+1)+"""</td>
                    <td><a href="""+url+""" target="_blank"><b>"""  +str(row['Title']) + """</b></a></td>
                    <td>"""+str(row['Score'])+"""</td>
                    <td>"""+str(row['TotalReplies'])+"""</td>
                    </tr>"""
    hs += "</table>"
    display(HTML(hs))

for ind, comp_row in competitions_df.iterrows():
    render_html_for_comp(comp_row["ForumId"], comp_row["Id"], comp_row["Title"], comp_row["Slug"], comp_row["Subtitle"], 12)



In [ ]: