Algorithm tweak

[mulundapm]

Investigate how the algorithm can be improved

Suggestions:
Currently interaction score is undermined by other parameters even though it has the highest weighting. Apply (and test if possible) a new algorithm where there is an adjustment on weighting.

 ((`normalized_source`.`avg_max_fraction`*3) + (`normalized_source`.`normalized_avg_watchtime`*1) + (coalesce(`interaction_count`.`interaction_user_count`/`normalized_source`.`user_count`,0)*16)+(`normalized_source`.`percentage_of_complete_viewers`*6)) AS `score_new`,
    ((`normalized_source`.`avg_max_fraction`*3) + (`normalized_source`.`normalized_avg_watchtime`*1) + (coalesce(`interaction_count`.`interaction_user_count`/`normalized_source`.`user_count`,0)*4)+(`normalized_source`.`percentage_of_complete_viewers`*2)) AS `score`