1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.

Building high quality account lists to follow by using advanced filters

Discussion in 'Twitter' started by kctnky, Jun 1, 2016.

  1. kctnky

    kctnky Registered Member

    Joined:
    Dec 26, 2014
    Messages:
    65
    Likes Received:
    15
    I am thinking to build accounts lists to follow at first. Then I will load them to scraped users list in followliker so that my accounts will follow them. I am hoping that I will increase the followback ratio with this method. My process to build such lists will be just like this;

    I will find similar accounts in my niche whose following and follower counts are both high. Such accounts most probably gain followers via the search/follow method same as like me. So, those followers have high potentials to follow back. But some people just buy fake followers or use #autoollowback like hashtags at early stages of accounts in order to boost them. It is also possible that some of those followers may be nothing more than thrash. What I mean by thrash is that they may be irrelevant to my niche or they may be from irrelevant (undesired) countries. Most importantly, they may be inactive for very long times. I just want to filter those accounts from my scraped users list in order not to spend my follows to them.

    So, I need a filtering algorithm which will have many steps. Some of these steps are just obvious and basic like checking following/follower ratio, last tweet time, etc. Some of automation bots on the shelf have already these features. However, I want advanced features. In this thread, I want you to write your reviews based on your experience. What points should I check in this filter? How can I sense and eliminate thrash accounts? How twitter security guys do this filtering so they can suspend us anytime they want? :)

    You can write your opinions and features that you want in this filtering algorithm even they are very extreme. Just let us discuss the algorithmic complexity of implementation of that feature. When I finish the first version you can be sure that I will share it with the contributors.
     
  2. bArAsc

    bArAsc Regular Member

    Joined:
    Jun 2, 2015
    Messages:
    204
    Likes Received:
    70
    Gender:
    Male
    Good options for bots would be scrape users with out links in bio cuz only famous people and botting accs use them for make money and usually no one will follow back from this accs
     
    • Thanks Thanks x 2
  3. AntonioSeegars1

    AntonioSeegars1 BANNED BANNED

    Joined:
    Apr 13, 2015
    Messages:
    61
    Likes Received:
    5
    DO you want link building service?
     
  4.  ﴾͡๏̯͡๏﴿.tk

     ﴾͡๏̯͡๏﴿.tk Senior Member

    Joined:
    Oct 6, 2010
    Messages:
    849
    Likes Received:
    157
    Occupation:
    ﴾͡๏̯͡๏﴿ ───█ ﴾͡●
    Location:
    twitter.com/SeX#.﴾͡๏̯͡๏﴿.
    Home Page:
    i compiled 2 "lists" with a high FollowBack ratio

    The used filters: •flwr/flwng ratio 90-110 •tweeted within 7 days •> 2.5k flwr

    Advantages:

    •boost up your acc to 10k flwr to be able to follow the max of 1k/day (in your niche)
    •probably the fastest way to get ~35 followers for free

    http://twitter.com/f4flist/following#__~67.3K
    what you can expect by following ↑
    [​IMG]
    created out of the 1st acc, so the "better LIST"
    http://twitter.com/followratio/following#__22k_target_50k

    BE AWARE:
    •the UNflw4unFLW ratio is as high as the FB ratio
    •to follow both accs makes less sense
     
    Last edited: Jun 2, 2016
  5. blackbadger

    blackbadger Registered Member

    Joined:
    Sep 11, 2013
    Messages:
    58
    Likes Received:
    13
    Location:
    UK
    I'm rethinking - I wrote about it in my journey thread - how I do this too after reading @DesignCamp post about it.

    I'm also thinking about building in filters like:

    ignore if They don't have a defaut profile
    ignore if They don't have a default profile image

    They MUST have a bio (filter out the cam girl, dating shit)
    They MUST have tweeted in the last 7 days
    They MUST have a follower/following ratio of 25% or less - not sure about this yet though ...
     
  6. kctnky

    kctnky Registered Member

    Joined:
    Dec 26, 2014
    Messages:
    65
    Likes Received:
    15
    Currently I am planning to design my filter with following steps. They will be all parametric and optional so that I will be able to adjust my filter with different settings. If you come with a new step to add this list I am always open to new ideas. You can also propose optimal values for these steps based on your experience. At the end of the day, I want to obtain a list of quality accounts having high follow back ratio. I'd prefer 10k high quality followers rather than 50k randomly profiled followers.

    - Follower count upper & lower limits
    - Following count upper & lower limits
    - Following number to follower number ratio
    - Tweets count upper & lower limits
    - Likes count upper & lower limits
    - Bio check (set, not set, having link)
    - Location
    - Last tweet time
    - Number of tweets posted in this week
    - Number of tweets posted in this month
    - Ratio of tweets containing links
    - Number of different domains used in links
    - Restriced keywords in tweets (such as #followback)
    - How many of followers it follows in fact (in ratio)

    As a level-end-boss;
    - Time histogram of posted tweets inside a day (I will use this data to derive time patterns for accounts to capture them online. Then I will group them based on this information).

    Btw, I am also looking for ways to derive user habits and interests from social media posts by using some kind of neural networks. It is actually an academic research in my education life rather than online marketing needs. If I can manage this research then 6-digits and an academic title wait for me. Wish me luck :)
     
    • Thanks Thanks x 1
    Last edited: Jun 3, 2016
  7. Olovarno

    Olovarno Registered Member

    Joined:
    Jan 19, 2010
    Messages:
    90
    Likes Received:
    12
    If you want to identify habits and interests, you should check about sentiment analysis and data mining. There are some papers about it already, so they might help you on it.
     
    • Thanks Thanks x 1
  8. blackbadger

    blackbadger Registered Member

    Joined:
    Sep 11, 2013
    Messages:
    58
    Likes Received:
    13
    Location:
    UK
    do you put any value on the presence/absence of default profile and default profile image?

    I'd be interested to see how you go with the NN stuff, I tried it 20+ years ago with horse racing but never got far ... reckon NN software has changed a lot since then though.