主成分得分-线性回归
通过主成分分析，采取降维的方法，最终选取了一个能解释收视情况76.62%的主成分，计算该用户每日收视综合得分。然后进行线性回归可以得出该用户的一个k值。
s_date | s_count | s_timelen | s_usecount | 综合得分 |
20140801 | 6 | 250 | 10 | -0.61 |
20140802 | 3 | 123 | 4 | -1.89 |
20140803 | 5 | 204 | 9 | -0.97 |
20140804 | 14 | 316 | 19 | 1.13 |
20140805 | 22 | 226 | 25 | 2.00 |
20140806 | 3 | 117 | 5 | -1.83 |
20140807 | 10 | 155 | 13 | -0.37 |
20140808 | 21 | 198 | 25 | 1.80 |
20140809 | 20 | 341 | 26 | 2.36 |
20140810 | 19 | 246 | 23 | 1.64 |
20140811 | 11 | 411 | 18 | 1.15 |
20140812 | 9 | 295 | 14 | 0.18 |
20140813 | 10 | 233 | 15 | 0.11 |
20140814 | 9 | 339 | 15 | 0.43 |
20140815 | 12 | 382 | 18 | 1.13 |
20140816 | 6 | 130 | 9 | -1.17 |
20140817 | 10 | 402 | 16 | 0.86 |
20140818 | 5 | 97 | 7 | -1.56 |
20140819 | 9 | 108 | 12 | -0.73 |
20140820 | 13 | 180 | 16 | 0.25 |
20140821 | 4 | 163 | 6 | -1.47 |
20140822 | 7 | 280 | 11 | -0.32 |
20140823 | 9 | 160 | 12 | -0.53 |
20140824 | 8 | 239 | 11 | -0.39 |
20140825 | 4 | 175 | 6 | -1.42 |
20140826 | 7 | 256 | 12 | -0.33 |
20140827 | 15 | 310 | 20 | 1.28 |
20140828 | 7 | 290 | 13 | -0.11 |
20140829 | 9 | 277 | 14 | 0.10 |
20140830 | 12 | 199 | 15 | 0.15 |
20140831 | 6 | 197 | 9 | -0.90 |
Variable | Coefficient | Std. Error | t-Statistic | Prob. |
C | 0.310194 | 0.429524 | 0.722179 | 0.4760 |
X | -0.019448 | 0.023432 | -0.829941 | 0.4134 |
R-squared | 0.023201 | Mean dependent var | -0.000968 | |
Adjusted R-squared | -0.010482 | S.D. dependent var | 1.160860 | |
S.E. of regression | 1.166928 | Akaike info criterion | 3.208967 | |
Sum squared resid | 39.48991 | Schwarz criterion | 3.301483 | |
Log likelihood | -47.73900 | Hannan-Quinn criter. | 3.239125 | |
F-statistic | 0.688802 | Durbin-Watson stat | 1.543558 | |
Prob(F-statistic) | 0.413352 | |||
样本中的拟合优度很低，显著性水平F值0.689，大于0.05，模型显著性不强。
收视时长当期与均值的比值
收视均值：235.4516129
8月31日收视时长：197
P=197/235.4516129=0.836689958
次均收视时长=总收视时长/总收视次数
L=7299/305=23.93114754
综合比值=当期之和/基期之和=（9+12+6）/（6+3+5）=1.928571429
