5. Conclusion

• Study the sensitivity of best precision+recall rate to the number of trades

• Given one set of local optimal parameters, study the sensitivity of precision

We have removed l∈ 0; 1000; 2000; 3000 number of bars to study the sensitivity in the two previous cases, which corresponds to several hours of trading data removed. Indeed one does not want to erase first flash crash detected in the data set and erase more buckets than the average prediction length to detect it. Moreover we

We summarize in Table 20 for each bar price structure the average percentage

change of local new best precision+recall rates with the number of bar erased.

• Its amplitude is not very big, at least for best precision+recall rate, as the

• Median bar price structure is far less sensitive than other price structure.

Bar price structure 1000 bars erased 2000 bars erased 3000 bars erased Last 3.089 2.166 0.939 First 2.410 3.649 6.727 Median 0.611 0.801 0.781 Mean 1.348 2.149 3.944

Average absolute percentage change of local best precision+recall rates with the number of bar erased for each

Bar price structure 1000 bars erased 2000 bars erased 3000 bars erased

Average absolute percentage change of local best precision+recall rates with the number of bar erased for each

Last 1.192 1.171 1.067 First 1.612 1.725 1.049 Median 3.514 3.137 1.396 Mean 2.648 3.180 2.489

• The sensitivity mentioned by Bodarenko and Anderson does exist.

and recall rates to bar price option and data removed.

would like to study to which extent VPIN is locally sensitive.

erased and to the bar price option.

Advanced Analytics and Artificial Intelligence Applications

4.2 Summary of results

Table 20.

Table 21.

70

bar price structure.

bar price structure.

4.2.1 Sensitivity of precision+recall rate

maximum change is about 6%.

We remark the following:

In this last section, we present first a general summary of our findings. Then we propose new suggestion of research concerning this precise subject.
