Sentiment Analysis on Mobile App Games
0 likes179 views
This document analyzed sentiment data from comments on three mobile games: Farmville Tropic Escape, Hit It Rich, and Poker. For each game, it identified the top words used, the distribution of positive, neutral, and negative sentiment scores for sample comments, and time periods with the most user engagement based on likes and comment threads. Recommendations for each game included fixing bugs, adjusting difficulty levels, highlighting popular features, and focusing customer support and bonus offers around peak engagement hours. The analysis aimed to understand user sentiment and demands to improve product impact and business performance.
Sentiment Analysis on Mobile App Games
- 1. Paglinawan, Julie Fernandez, Rachel Gayo, Ivy Analyst Marketer Solitaire player Data Saiyans Tackling Facebook data for you
- 6. launched July 2007 300 M users
- 10. Why look into Sentiments? To make sure that the business is built for its customers.
- 11. To be able to identify the demand and fuss in the market in creating stronger product impact.
- 12. ● What is the sentiment per comment and game? ● When is the best time to post? ● What are the most commonly used words? ● What do users want?
- 14. Likes and Shares correlation per Game
- 15. Taking a look at the Data
- 16. Taking a look at the Data Count of Reactions and Comments per Game
- 18. The challenges in the Data Imbalanced Data set Balanced Data set Comment Thread Distribution
- 19. The challenges in the Data How it was solved: By applying SMOTE( Synthetic Minority Over-Sampling Technique)
- 21. The challenges in the Data Handling comments that are not in English How it was solved: By using the corpus “words “ which contains the list of English words. nltk.corpus.words Non english words are then dropped.
- 22. Machines doing the heavy Work 87% accuracy score from the balanced data set 39%Average Precision Results from Random Forest Classification only
- 23. TOP 20 Most Important Features Out of 543features made from vectorization and Sentiment Intensity Analyzer, here are the most Important features that help explain our model.
- 24. What’s in it for the
- 25. Compound SCORES ● 47.4% Positive ● 36.6% Neutral ● 16.0% Negative ● 73.2% Positive ● 20.5% Neutral ● 6.3% Negative ● 49.2% Positive ● 38.5% Neutral ● 12.3% Negative
- 26. Top 20 most talked about Farmville Tropic Escape most common words
- 27. FVTE, Score Distribution per Comment Comment Message Pos Neu Neg Likes the friendships i have built in this game is amazing! our club is a big family! thank you for bringing strangers together to enjoy a fun game 0.63 0.37 0.00 0 why is my cruise say it’s loading but .... no cruise ship 0.00 1.00 0.00 2 never got did everything zynga recommended. oh well my club tells me it's not that fun anyway lol 0.37 0.48 0.15 10 zemi is a scammer , i log in every day and he never gives me the big gift whatever it is lol 0.00 0.77 0.23 8 miss my lighthouse. even if the landmarks are full time. earned it ?? 0.00 0.76 0.24 5 farmville: tropic escape i've earned the extra 5 days! woohoo!! 0.56 0.44 0.00 5
- 28. FVTE, Sentiment Deep Dive Generally we get more Engagement in the afternoon to early in the evening.
- 29. Top 20 most talked about Hit It Rich's most common words
- 30. HIT, Score Distribution per Comment Comment Message Pos Neu Neg Likes this is a waste of time you can extend it till 2050 and we still want be able to complete it 0.28 0.46 0.26 27 thank you for the brkfst coins. 0.71 0.29 0.00 1 please give me back my old hit it rich on facebook thank you 0.68 0.32 0.00 11 we would enjoy it more if we could get some big hits and non duplicate cards. 0.29 0.71 0.00 9 because we are all gluten's for punishment. 0.00 0.24 0.76 0 btw...i will never play your slot ventures again 0.00 1.33 0.67 0 i've lost quite a bit today to and haven't moved once on path to riches. ðÿ™„ðÿ™„ 0.27 0.53 0.20 0
- 31. HIR, Sentiment Deep Dive 2AM, 7PM and 8PM are the significant hours with the most engagement
- 32. Top 20 most talked aboutPOKER's most common words
- 33. POKER, Score Distribution per Comment Comment Message Pos Neu Neg Likes hand one with with 3 pairs 10 kicker 0.52 0.48 0.00 56 bots will bet heavily knowing the river card. they are thick in zynga. 0.00 1.00 0.00 14 let's play some poker! love playing against people that have no clue! ðÿ¤£ 0.64 0.36 0.00 9 straight flush beats any full house. 0.64 1.36 0.00 0 go in hell corupted mafia 0.00 0.18 0.82 0 stopped playing to much cheating going on 0.24 0.20 0.57 1 but is it really a bad beat when the hands are predetermined? 0.34 0.29 0.37 0 if hand 2 makes any bet at the flop, turn, or river, they win. no way hand 1 calls. 0.55 0.32 0.13 1 result is good i won lots hands 0.75 0.25 0.00 2
- 34. POKER, Sentiment Deep Dive 5PM has the most engagement about 301 likes and 62 comment threads.
- 37. RECOMMENDATION FVTE ● Fix app Bugs. ● Less Game Difficulty ● Less Game Goals ● High chances of user engagement every 3PM to 8PM ● Slight increase in positive Sentiment trend on April 20 when page owners posts about lava rocks, cruise and how users can meet on other players.
- 38. RECOMMENDATION HIR ● Breakfast coins ● Bring back classic version ● Getting the new card ● Freaki tiki feature ● Load time ● HIT peak hour is every 8PM for user enaggement
- 39. RECOMMENDATION POKER ● Fix game bugs/bots ● Provide quick and efficient customer service for missing chips/tokens ● People are looking forward to friday free tokens/chips ● Bring back the classic version ● High chances of user engagement every 5 pm